{"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "The lecture will start at 21:50 in room B2-14."}, {"role": "assistant", "content": "The lecture will start at twenty-one fifty in room B two fourteen."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit www.correo.es or contact us at 018-890 7982."}, {"role": "assistant", "content": "To reset your password, please visit w w w dot correo dot es or contact us at zero one eight eight nine zero seven nine eight two."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your OTP is 22108, valid for 28 minutes."}, {"role": "assistant", "content": "Your OTP is two two one zero eight, valid for twenty-eight minutes."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "The lecture will start at 9:00 a.m. in room A12."}, {"role": "assistant", "content": "The lecture will start at nine a m in room A twelve."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Please contact us at 019-027 5460 if you need to reschedule."}, {"role": "assistant", "content": "Please contact us at zero one nine zero two seven five four six zero if you need to reschedule."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Please contact us at 1-300-88-0391 if you need to reschedule."}, {"role": "assistant", "content": "Please contact us at one three zero zero eight eight zero three nine one if you need to reschedule."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://www.correo.es/support or contact us at +60 11-377 0771."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash www dot correo dot es slash support or contact us at plus six zero one one three seven seven zero seven seven one."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://mail.fr/help and enter your student email pedro@posta.it."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash mail dot fr slash help and enter your student email pedro at posta dot it."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Please contact us at 017-99705228 if you need to reschedule."}, {"role": "assistant", "content": "Please contact us at zero one seven nine nine seven zero five two two eight if you need to reschedule."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://yahoo.com.ph/faq and enter your student email jan.devries@scicom.com.my."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash yahoo dot com dot ph slash faq and enter your student email jan dot devries at scicom dot com dot my."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit www.bank.my/login and enter your student email jan.devries@yahoo.com."}, {"role": "assistant", "content": "To reset your password, please visit w w w dot bank dot my slash login and enter your student email jan dot devries at yahoo dot com."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your OTP is 2416, valid for 12 minutes."}, {"role": "assistant", "content": "Your OTP is two four one six, valid for twelve minutes."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your OTP is 644748, valid for 790 minutes."}, {"role": "assistant", "content": "Your OTP is six four four seven four eight, valid for seven hundred ninety minutes."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://www.company.co/support and enter your student email kasia@posta.it."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash www dot company dot co slash support and enter your student email kasia at posta dot it."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Please contact us at 03-9806 4947 if you need to reschedule."}, {"role": "assistant", "content": "Please contact us at zero three nine eight zero six four nine four seven if you need to reschedule."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://www.web.de/support or contact us at 03-8370 9898."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash www dot web dot de slash support or contact us at zero three eight three seven zero nine eight nine eight."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://web.de/faq and enter your student email maria@sapo.pt."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash web dot de slash faq and enter your student email maria at sapo dot pt."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your OTP is 214896, valid for 556 minutes."}, {"role": "assistant", "content": "Your OTP is two one four eight nine six, valid for five hundred fifty-six minutes."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your OTP is 61513, valid for 819 minutes."}, {"role": "assistant", "content": "Your OTP is six one five one three, valid for eight hundred nineteen minutes."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Please contact us at 1-300-88-7683 if you need to reschedule."}, {"role": "assistant", "content": "Please contact us at one three zero zero eight eight seven six eight three if you need to reschedule."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://correo.es or contact us at 014-49597548."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash correo dot es or contact us at zero one four four nine five nine seven five four eight."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://web.de or contact us at +60 15-647 5806."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash web dot de or contact us at plus six zero one five six four seven five eight zero six."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your OTP is 262959, valid for 1 minutes."}, {"role": "assistant", "content": "Your OTP is two six two nine five nine, valid for one minutes."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://posta.it/faq and enter your student email help@bank.my."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash posta dot it slash faq and enter your student email help at bank dot my."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "The lecture will start at 2pm in room WXY 1234."}, {"role": "assistant", "content": "The lecture will start at two p m in room WXY one two three four."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Please contact us at +60 18-348 1084 if you need to reschedule."}, {"role": "assistant", "content": "Please contact us at plus six zero one eight three four eight one zero eight four if you need to reschedule."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit www.ziggo.nl or contact us at 012-28873439."}, {"role": "assistant", "content": "To reset your password, please visit w w w dot ziggo dot nl or contact us at zero one two two eight eight seven three four three nine."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "The lecture will start at 10.45PM in room WXY 1234."}, {"role": "assistant", "content": "The lecture will start at ten forty-five P M in room WXY one two three four."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://correo.es or contact us at 03-6714 6437."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash correo dot es or contact us at zero three six seven one four six four three seven."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit www.correo.es and enter your student email help@yahoo.com."}, {"role": "assistant", "content": "To reset your password, please visit w w w dot correo dot es and enter your student email help at yahoo dot com."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Please contact us at 015-18034462 if you need to reschedule."}, {"role": "assistant", "content": "Please contact us at zero one five one eight zero three four four six two if you need to reschedule."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your OTP is 88525, valid for 64 minutes."}, {"role": "assistant", "content": "Your OTP is eight eight five two five, valid for sixty-four minutes."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your OTP is 9870, valid for 9 minutes."}, {"role": "assistant", "content": "Your OTP is nine eight seven zero, valid for nine minutes."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "The lecture will start at 19:45 in room WXY 1234."}, {"role": "assistant", "content": "The lecture will start at nineteen forty-five in room WXY one two three four."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your OTP is 455404, valid for 40 minutes."}, {"role": "assistant", "content": "Your OTP is four five five four zero four, valid for forty minutes."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://gmail.co.id/faq or contact us at +60 12-074 1889."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash gmail dot co dot id slash faq or contact us at plus six zero one two zero seven four one eight eight nine."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://yahoo.com/faq or contact us at 011-483 3031."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash yahoo dot com slash faq or contact us at zero one one four eight three three zero three one."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Please contact us at 015-13227108 if you need to reschedule."}, {"role": "assistant", "content": "Please contact us at zero one five one three two two seven one zero eight if you need to reschedule."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Please contact us at 016-70940446 if you need to reschedule."}, {"role": "assistant", "content": "Please contact us at zero one six seven zero nine four zero four four six if you need to reschedule."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit www.yahoo.com.ph or contact us at +60 16-033 6736."}, {"role": "assistant", "content": "To reset your password, please visit w w w dot yahoo dot com dot ph or contact us at plus six zero one six zero three three six seven three six."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://www.sapo.pt/support or contact us at +60 18-794 1613."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash www dot sapo dot pt slash support or contact us at plus six zero one eight seven nine four one six one three."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "The lecture will start at 9.45pm in room REF-2024-0113."}, {"role": "assistant", "content": "The lecture will start at nine forty-five p m in room REF two zero two four zero one one three."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://mail.fr/faq and enter your student email luca92@mail.fr."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash mail dot fr slash faq and enter your student email luca nine two at mail dot fr."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://ziggo.nl/help and enter your student email info@sapo.pt."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash ziggo dot nl slash help and enter your student email info at sapo dot pt."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Please contact us at +60 13-616 9887 if you need to reschedule."}, {"role": "assistant", "content": "Please contact us at plus six zero one three six one six nine eight eight seven if you need to reschedule."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit www.sltnet.lk/login or contact us at 011-122 3218."}, {"role": "assistant", "content": "To reset your password, please visit w w w dot sltnet dot lk slash login or contact us at zero one one one two two three two one eight."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://outlook.sa or contact us at 03-2344 5232."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash outlook dot sa or contact us at zero three two three four four five two three two."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your OTP is 8327, valid for 1 minutes."}, {"role": "assistant", "content": "Your OTP is eight three two seven, valid for one minutes."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "The lecture will start at 9PM in room ABC1234."}, {"role": "assistant", "content": "The lecture will start at nine P M in room ABC one two three four."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your OTP is 91945, valid for 9 minutes."}, {"role": "assistant", "content": "Your OTP is nine one nine four five, valid for nine minutes."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "The lecture will start at 2:30 in room INV0472."}, {"role": "assistant", "content": "The lecture will start at two thirty in room INV zero four seven two."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://www.yahoo.com.ph/support and enter your student email help62@web.de."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash www dot yahoo dot com dot ph slash support and enter your student email help six two at web dot de."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://gmail.co.id/help or contact us at 019-92425914."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash gmail dot co dot id slash help or contact us at zero one nine nine two four two five nine one four."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://onet.pl/faq or contact us at 1-300-88-4157."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash onet dot pl slash faq or contact us at one three zero zero eight eight four one five seven."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://sltnet.lk and enter your student email info@ziggo.nl."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash sltnet dot lk and enter your student email info at ziggo dot nl."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your OTP is 479291, valid for 183 minutes."}, {"role": "assistant", "content": "Your OTP is four seven nine two nine one, valid for one hundred eighty-three minutes."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your OTP is 5317, valid for 5 minutes."}, {"role": "assistant", "content": "Your OTP is five three one seven, valid for five minutes."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit www.gmail.com/login and enter your student email kasia@yahoo.com.ph."}, {"role": "assistant", "content": "To reset your password, please visit w w w dot gmail dot com slash login and enter your student email kasia at yahoo dot com dot ph."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://mail.fr/help and enter your student email juan75@example.com."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash mail dot fr slash help and enter your student email juan seven five at example dot com."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit www.bank.my and enter your student email chen8872@onet.pl."}, {"role": "assistant", "content": "To reset your password, please visit w w w dot bank dot my and enter your student email chen eight eight seven two at onet dot pl."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Please contact us at 03-1421 1165 if you need to reschedule."}, {"role": "assistant", "content": "Please contact us at zero three one four two one one one six five if you need to reschedule."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit www.gmail.com and enter your student email siti@onet.pl."}, {"role": "assistant", "content": "To reset your password, please visit w w w dot gmail dot com and enter your student email siti at onet dot pl."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your OTP is 64467, valid for 2 minutes."}, {"role": "assistant", "content": "Your OTP is six four four six seven, valid for two minutes."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "The lecture will start at 12:10 in room TK-88213."}, {"role": "assistant", "content": "The lecture will start at twelve ten in room TK eight eight two one three."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Please contact us at 1-300-88-7695 if you need to reschedule."}, {"role": "assistant", "content": "Please contact us at one three zero zero eight eight seven six nine five if you need to reschedule."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://www.correo.es/support and enter your student email anna@outlook.sa."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash www dot correo dot es slash support and enter your student email anna at outlook dot sa."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your OTP is 68180, valid for 8 minutes."}, {"role": "assistant", "content": "Your OTP is six eight one eight zero, valid for eight minutes."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your OTP is 78049, valid for 557 minutes."}, {"role": "assistant", "content": "Your OTP is seven eight zero four nine, valid for five hundred fifty-seven minutes."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://yahoo.com.ph or contact us at 1-300-88-1508."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash yahoo dot com dot ph or contact us at one three zero zero eight eight one five zero eight."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Please contact us at 1-300-88-0411 if you need to reschedule."}, {"role": "assistant", "content": "Please contact us at one three zero zero eight eight zero four one one if you need to reschedule."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "The lecture will start at 1am in room REF-2024-0113."}, {"role": "assistant", "content": "The lecture will start at one a m in room REF two zero two four zero one one three."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://ziggo.nl/faq or contact us at 012-520 1025."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash ziggo dot nl slash faq or contact us at zero one two five two zero one zero two five."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Please contact us at 011-22973828 if you need to reschedule."}, {"role": "assistant", "content": "Please contact us at zero one one two two nine seven three eight two eight if you need to reschedule."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Please contact us at 03-7189 0737 if you need to reschedule."}, {"role": "assistant", "content": "Please contact us at zero three seven one eight nine zero seven three seven if you need to reschedule."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your OTP is 76698, valid for 766 minutes."}, {"role": "assistant", "content": "Your OTP is seven six six nine eight, valid for seven hundred sixty-six minutes."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your OTP is 5283, valid for 367 minutes."}, {"role": "assistant", "content": "Your OTP is five two eight three, valid for three hundred sixty-seven minutes."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit www.sltnet.lk/login or contact us at 03-3872 5611."}, {"role": "assistant", "content": "To reset your password, please visit w w w dot sltnet dot lk slash login or contact us at zero three three eight seven two five six one one."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "The lecture will start at 10:20 p.m. in room REF-2024-0113."}, {"role": "assistant", "content": "The lecture will start at ten twenty p m in room REF two zero two four zero one one three."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://correo.es and enter your student email info25@onet.pl."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash correo dot es and enter your student email info two five at onet dot pl."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Please contact us at 03-4905 1486 if you need to reschedule."}, {"role": "assistant", "content": "Please contact us at zero three four nine zero five one four eight six if you need to reschedule."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://www.bank.my/support and enter your student email ravi41@web.de."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash www dot bank dot my slash support and enter your student email ravi four one at web dot de."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "The lecture will start at 8:10 am in room X-12340567."}, {"role": "assistant", "content": "The lecture will start at eight ten a m in room X one two three four zero five six seven."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your OTP is 224018, valid for 9 minutes."}, {"role": "assistant", "content": "Your OTP is two two four zero one eight, valid for nine minutes."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Please contact us at 1-300-88-1960 if you need to reschedule."}, {"role": "assistant", "content": "Please contact us at one three zero zero eight eight one nine six zero if you need to reschedule."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit www.scicom.com.my/login or contact us at 016-297 0713."}, {"role": "assistant", "content": "To reset your password, please visit w w w dot scicom dot com dot my slash login or contact us at zero one six two nine seven zero seven one three."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://www.outlook.sa/support or contact us at 012-09652985."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash www dot outlook dot sa slash support or contact us at zero one two zero nine six five two nine eight five."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://www.sltnet.lk/support and enter your student email nurul_h@correo.es."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash www dot sltnet dot lk slash support and enter your student email nurul underscore h at correo dot es."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "The lecture will start at 8:10 p.m. in room REF-2024-0113."}, {"role": "assistant", "content": "The lecture will start at eight ten p m in room REF two zero two four zero one one three."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "The lecture will start at 10.45AM in room TK-88213."}, {"role": "assistant", "content": "The lecture will start at ten forty-five A M in room TK eight eight two one three."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Please contact us at 010-25719689 if you need to reschedule."}, {"role": "assistant", "content": "Please contact us at zero one zero two five seven one nine six eight nine if you need to reschedule."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit www.example.com and enter your student email omar@correo.es."}, {"role": "assistant", "content": "To reset your password, please visit w w w dot example dot com and enter your student email omar at correo dot es."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit www.outlook.sa and enter your student email billing@sapo.pt."}, {"role": "assistant", "content": "To reset your password, please visit w w w dot outlook dot sa and enter your student email billing at sapo dot pt."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://scicom.com.my/faq or contact us at 03-0108 9489."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash scicom dot com dot my slash faq or contact us at zero three zero one zero eight nine four eight nine."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://outlook.sa/help and enter your student email anna88@yahoo.com."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash outlook dot sa slash help and enter your student email anna eight eight at yahoo dot com."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit www.gmail.co.id/login or contact us at 019-233 9971."}, {"role": "assistant", "content": "To reset your password, please visit w w w dot gmail dot co dot id slash login or contact us at zero one nine two three three nine nine seven one."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit www.bank.my and enter your student email juan@sapo.pt."}, {"role": "assistant", "content": "To reset your password, please visit w w w dot bank dot my and enter your student email juan at sapo dot pt."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "The lecture will start at 7.45pm in room AK6120."}, {"role": "assistant", "content": "The lecture will start at seven forty-five p m in room AK six one two zero."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Please contact us at 018-96964576 if you need to reschedule."}, {"role": "assistant", "content": "Please contact us at zero one eight nine six nine six four five seven six if you need to reschedule."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://www.onet.pl/support or contact us at 1-300-88-3878."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash www dot onet dot pl slash support or contact us at one three zero zero eight eight three eight seven eight."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "The lecture will start at 15:00 in room ABC1234."}, {"role": "assistant", "content": "The lecture will start at fifteen o'clock in room ABC one two three four."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Please contact us at 016-94868152 if you need to reschedule."}, {"role": "assistant", "content": "Please contact us at zero one six nine four eight six eight one five two if you need to reschedule."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Please contact us at 03-2300 8832 if you need to reschedule."}, {"role": "assistant", "content": "Please contact us at zero three two three zero zero eight eight three two if you need to reschedule."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://posta.it or contact us at 03-4703 5399."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash posta dot it or contact us at zero three four seven zero three five three nine nine."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://company.co/help and enter your student email jan.devries@gmail.co.id."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash company dot co slash help and enter your student email jan dot devries at gmail dot co dot id."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your OTP is 627098, valid for 224 minutes."}, {"role": "assistant", "content": "Your OTP is six two seven zero nine eight, valid for two hundred twenty-four minutes."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://gmail.co.id or contact us at 010-06379010."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash gmail dot co dot id or contact us at zero one zero zero six three seven nine zero one zero."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://yahoo.com.ph/faq or contact us at 1-300-88-1175."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash yahoo dot com dot ph slash faq or contact us at one three zero zero eight eight one one seven five."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit www.sapo.pt/login and enter your student email dilani@gmail.co.id."}, {"role": "assistant", "content": "To reset your password, please visit w w w dot sapo dot pt slash login and enter your student email dilani at gmail dot co dot id."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your OTP is 72498, valid for 200 minutes."}, {"role": "assistant", "content": "Your OTP is seven two four nine eight, valid for two hundred minutes."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your OTP is 6134, valid for 735 minutes."}, {"role": "assistant", "content": "Your OTP is six one three four, valid for seven hundred thirty-five minutes."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Please contact us at +60 18-467 8538 if you need to reschedule."}, {"role": "assistant", "content": "Please contact us at plus six zero one eight four six seven eight five three eight if you need to reschedule."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your OTP is 1935, valid for 29 minutes."}, {"role": "assistant", "content": "Your OTP is nineteen thirty-five, valid for twenty-nine minutes."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://posta.it/help or contact us at 011-407 7378."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash posta dot it slash help or contact us at zero one one four zero seven seven three seven eight."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit https://bank.my/help or contact us at 1-300-88-9028."}, {"role": "assistant", "content": "To reset your password, please visit h t t p s colon slash slash bank dot my slash help or contact us at one three zero zero eight eight nine zero two eight."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Please contact us at 017-94941436 if you need to reschedule."}, {"role": "assistant", "content": "Please contact us at zero one seven nine four nine four one four three six if you need to reschedule."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "The lecture will start at 9:50 AM in room ABC1234."}, {"role": "assistant", "content": "The lecture will start at nine fifty A M in room ABC one two three four."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your OTP is 4545, valid for 84 minutes."}, {"role": "assistant", "content": "Your OTP is four five four five, valid for eighty-four minutes."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit www.yahoo.com.ph or contact us at 03-4311 1602."}, {"role": "assistant", "content": "To reset your password, please visit w w w dot yahoo dot com dot ph or contact us at zero three four three one one one six zero two."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "The lecture will start at 3a.m. in room A12."}, {"role": "assistant", "content": "The lecture will start at three a m in room A twelve."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "To reset your password, please visit www.sapo.pt or contact us at 010-91188527."}, {"role": "assistant", "content": "To reset your password, please visit w w w dot sapo dot pt or contact us at zero one zero nine one one eight eight five two seven."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "The lecture will start at 9.40a.m. in room WXY 1234."}, {"role": "assistant", "content": "The lecture will start at nine forty a m in room WXY one two three four."}], "lang": "en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah terpilih untuk menerima rebat sebanyak 56% bagi langganan tahun 1981."}, {"role": "assistant", "content": "Anda telah terpilih untuk menerima rebat sebanyak lima puluh enam peratus bagi langganan tahun seribu sembilan ratus lapan puluh satu."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah terpilih untuk menerima rebat sebanyak 57% bagi langganan tahun 1962."}, {"role": "assistant", "content": "Anda telah terpilih untuk menerima rebat sebanyak lima puluh tujuh peratus bagi langganan tahun seribu sembilan ratus enam puluh dua."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah dipilih sebagai calon ke-19 untuk temu duga pada 26/7/2006."}, {"role": "assistant", "content": "Anda telah dipilih sebagai calon kesembilan belas untuk temu duga pada dua puluh enam Julai dua ribu enam."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari 5 pagi hingga 3 petang pada 21-02-2007."}, {"role": "assistant", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari lima pagi hingga tiga petang pada dua puluh satu Februari dua ribu tujuh."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari 11.10 pagi hingga 9.15 pagi pada 06-04-2007."}, {"role": "assistant", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari sebelas sepuluh pagi hingga sembilan lima belas pagi pada enam April dua ribu tujuh."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari 4:15 petang hingga 22:15 pada 20 Jun."}, {"role": "assistant", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari empat lima belas petang hingga dua puluh dua lima belas pada dua puluh Jun."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah dipilih sebagai calon ke-4 untuk temu duga pada 20-06-2002."}, {"role": "assistant", "content": "Anda telah dipilih sebagai calon keempat untuk temu duga pada dua puluh Jun dua ribu dua."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda adalah pesakit ke-6 yang dijadualkan untuk sesi konsultasi hari ini."}, {"role": "assistant", "content": "Anda adalah pesakit keenam yang dijadualkan untuk sesi konsultasi hari ini."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah dipilih sebagai calon ke-27 untuk temu duga pada 9/10/2005."}, {"role": "assistant", "content": "Anda telah dipilih sebagai calon kedua puluh tujuh untuk temu duga pada sembilan Oktober dua ribu lima."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Barangan anda telah sampai di pusat agihan ke-9 sebelum dihantar ke rumah."}, {"role": "assistant", "content": "Barangan anda telah sampai di pusat agihan kesembilan sebelum dihantar ke rumah."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari 12 petang hingga 3 malam pada 20 September 2006."}, {"role": "assistant", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari dua belas petang hingga tiga malam pada dua puluh September dua ribu enam."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah dipilih sebagai calon ke-28 untuk temu duga pada 09-10-2023."}, {"role": "assistant", "content": "Anda telah dipilih sebagai calon kedua puluh lapan untuk temu duga pada sembilan Oktober dua ribu dua puluh tiga."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda adalah pesakit ke-19 yang dijadualkan untuk sesi konsultasi hari ini."}, {"role": "assistant", "content": "Anda adalah pesakit kesembilan belas yang dijadualkan untuk sesi konsultasi hari ini."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah dipilih sebagai calon ke-17 untuk temu duga pada 08-04-2000."}, {"role": "assistant", "content": "Anda telah dipilih sebagai calon ketujuh belas untuk temu duga pada lapan April dua ribu."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda adalah pesakit ke-15 yang dijadualkan untuk sesi konsultasi hari ini."}, {"role": "assistant", "content": "Anda adalah pesakit kelima belas yang dijadualkan untuk sesi konsultasi hari ini."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari 6:45 hingga 11.05 pagi pada 3/9/1998."}, {"role": "assistant", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari enam empat puluh lima hingga sebelas kosong lima pagi pada tiga September seribu sembilan ratus sembilan puluh lapan."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah terpilih untuk menerima rebat sebanyak 24.5% bagi langganan tahun 1952."}, {"role": "assistant", "content": "Anda telah terpilih untuk menerima rebat sebanyak dua puluh empat perpuluhan lima peratus bagi langganan tahun seribu sembilan ratus lima puluh dua."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah dipilih sebagai calon ke-14 untuk temu duga pada 28 Disember 2003."}, {"role": "assistant", "content": "Anda telah dipilih sebagai calon keempat belas untuk temu duga pada dua puluh lapan Disember dua ribu tiga."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah terpilih untuk menerima rebat sebanyak 35% bagi langganan tahun 1975."}, {"role": "assistant", "content": "Anda telah terpilih untuk menerima rebat sebanyak tiga puluh lima peratus bagi langganan tahun seribu sembilan ratus tujuh puluh lima."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari 6:00 pagi hingga 21:05 pada 13 Jun."}, {"role": "assistant", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari enam pagi hingga dua puluh satu kosong lima pada tiga belas Jun."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah terpilih untuk menerima rebat sebanyak 25% bagi langganan tahun 2023."}, {"role": "assistant", "content": "Anda telah terpilih untuk menerima rebat sebanyak dua puluh lima peratus bagi langganan tahun dua ribu dua puluh tiga."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Barangan anda telah sampai di pusat agihan ke-3 sebelum dihantar ke rumah."}, {"role": "assistant", "content": "Barangan anda telah sampai di pusat agihan ketiga sebelum dihantar ke rumah."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah terpilih untuk menerima rebat sebanyak 65% bagi langganan tahun 1960."}, {"role": "assistant", "content": "Anda telah terpilih untuk menerima rebat sebanyak enam puluh lima peratus bagi langganan tahun seribu sembilan ratus enam puluh."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Barangan anda telah sampai di pusat agihan ke-27 sebelum dihantar ke rumah."}, {"role": "assistant", "content": "Barangan anda telah sampai di pusat agihan kedua puluh tujuh sebelum dihantar ke rumah."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah dipilih sebagai calon ke-22 untuk temu duga pada 4/2/2006."}, {"role": "assistant", "content": "Anda telah dipilih sebagai calon kedua puluh dua untuk temu duga pada empat Februari dua ribu enam."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah terpilih untuk menerima rebat sebanyak 98% bagi langganan tahun 1977."}, {"role": "assistant", "content": "Anda telah terpilih untuk menerima rebat sebanyak sembilan puluh lapan peratus bagi langganan tahun seribu sembilan ratus tujuh puluh tujuh."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah terpilih untuk menerima rebat sebanyak 24.2% bagi langganan tahun 1973."}, {"role": "assistant", "content": "Anda telah terpilih untuk menerima rebat sebanyak dua puluh empat perpuluhan dua peratus bagi langganan tahun seribu sembilan ratus tujuh puluh tiga."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda adalah pesakit ke-28 yang dijadualkan untuk sesi konsultasi hari ini."}, {"role": "assistant", "content": "Anda adalah pesakit kedua puluh lapan yang dijadualkan untuk sesi konsultasi hari ini."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari 4 malam hingga 3:40 pada 2/9/2000."}, {"role": "assistant", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari empat malam hingga tiga empat puluh pada dua September dua ribu."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda adalah pesakit ke-2 yang dijadualkan untuk sesi konsultasi hari ini."}, {"role": "assistant", "content": "Anda adalah pesakit kedua yang dijadualkan untuk sesi konsultasi hari ini."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Barangan anda telah sampai di pusat agihan ke-21 sebelum dihantar ke rumah."}, {"role": "assistant", "content": "Barangan anda telah sampai di pusat agihan kedua puluh satu sebelum dihantar ke rumah."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari 11 malam hingga 2.00 malam pada 4/2/1995."}, {"role": "assistant", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari sebelas malam hingga dua malam pada empat Februari seribu sembilan ratus sembilan puluh lima."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah dipilih sebagai calon ke-6 untuk temu duga pada 16/12/2009."}, {"role": "assistant", "content": "Anda telah dipilih sebagai calon keenam untuk temu duga pada enam belas Disember dua ribu sembilan."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari 3.15 malam hingga 3 malam pada 17 Julai."}, {"role": "assistant", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari tiga lima belas malam hingga tiga malam pada tujuh belas Julai."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Barangan anda telah sampai di pusat agihan ke-5 sebelum dihantar ke rumah."}, {"role": "assistant", "content": "Barangan anda telah sampai di pusat agihan kelima sebelum dihantar ke rumah."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda adalah pesakit ke-25 yang dijadualkan untuk sesi konsultasi hari ini."}, {"role": "assistant", "content": "Anda adalah pesakit kedua puluh lima yang dijadualkan untuk sesi konsultasi hari ini."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah terpilih untuk menerima rebat sebanyak 15% bagi langganan tahun 1966."}, {"role": "assistant", "content": "Anda telah terpilih untuk menerima rebat sebanyak lima belas peratus bagi langganan tahun seribu sembilan ratus enam puluh enam."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda adalah pesakit ke-13 yang dijadualkan untuk sesi konsultasi hari ini."}, {"role": "assistant", "content": "Anda adalah pesakit ketiga belas yang dijadualkan untuk sesi konsultasi hari ini."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda adalah pesakit ke-18 yang dijadualkan untuk sesi konsultasi hari ini."}, {"role": "assistant", "content": "Anda adalah pesakit kelapan belas yang dijadualkan untuk sesi konsultasi hari ini."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah terpilih untuk menerima rebat sebanyak 54.3% bagi langganan tahun 1954."}, {"role": "assistant", "content": "Anda telah terpilih untuk menerima rebat sebanyak lima puluh empat perpuluhan tiga peratus bagi langganan tahun seribu sembilan ratus lima puluh empat."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda adalah pesakit ke-3 yang dijadualkan untuk sesi konsultasi hari ini."}, {"role": "assistant", "content": "Anda adalah pesakit ketiga yang dijadualkan untuk sesi konsultasi hari ini."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Barangan anda telah sampai di pusat agihan ke-23 sebelum dihantar ke rumah."}, {"role": "assistant", "content": "Barangan anda telah sampai di pusat agihan kedua puluh tiga sebelum dihantar ke rumah."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari 5.45 pagi hingga 5:20 petang pada 06-08-2017."}, {"role": "assistant", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari lima empat puluh lima pagi hingga lima dua puluh petang pada enam Ogos dua ribu tujuh belas."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah terpilih untuk menerima rebat sebanyak 7% bagi langganan tahun 1959."}, {"role": "assistant", "content": "Anda telah terpilih untuk menerima rebat sebanyak tujuh peratus bagi langganan tahun seribu sembilan ratus lima puluh sembilan."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Barangan anda telah sampai di pusat agihan ke-6 sebelum dihantar ke rumah."}, {"role": "assistant", "content": "Barangan anda telah sampai di pusat agihan keenam sebelum dihantar ke rumah."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda adalah pesakit ke-14 yang dijadualkan untuk sesi konsultasi hari ini."}, {"role": "assistant", "content": "Anda adalah pesakit keempat belas yang dijadualkan untuk sesi konsultasi hari ini."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah dipilih sebagai calon ke-6 untuk temu duga pada 26-07-2004."}, {"role": "assistant", "content": "Anda telah dipilih sebagai calon keenam untuk temu duga pada dua puluh enam Julai dua ribu empat."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari 20:00 hingga 10:05 pagi pada 27-12-2022."}, {"role": "assistant", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari dua puluh hingga sepuluh kosong lima pagi pada dua puluh tujuh Disember dua ribu dua puluh dua."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari 23:00 hingga 9.00 malam pada 02-07-2027."}, {"role": "assistant", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari dua puluh tiga hingga sembilan malam pada dua Julai dua ribu dua puluh tujuh."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari 1 malam hingga 4:05 pada 23-11-1996."}, {"role": "assistant", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari satu malam hingga empat kosong lima pada dua puluh tiga November seribu sembilan ratus sembilan puluh enam."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari 1 malam hingga 8:10 pada 09-07-2004."}, {"role": "assistant", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari satu malam hingga lapan sepuluh pada sembilan Julai dua ribu empat."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah dipilih sebagai calon ke-13 untuk temu duga pada 04-12-2026."}, {"role": "assistant", "content": "Anda telah dipilih sebagai calon ketiga belas untuk temu duga pada empat Disember dua ribu dua puluh enam."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah terpilih untuk menerima rebat sebanyak 28% bagi langganan tahun 1974."}, {"role": "assistant", "content": "Anda telah terpilih untuk menerima rebat sebanyak dua puluh lapan peratus bagi langganan tahun seribu sembilan ratus tujuh puluh empat."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari 7 malam hingga 5 pagi pada 20 Jun 2023."}, {"role": "assistant", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari tujuh malam hingga lima pagi pada dua puluh Jun dua ribu dua puluh tiga."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah dipilih sebagai calon ke-24 untuk temu duga pada 7 Mei."}, {"role": "assistant", "content": "Anda telah dipilih sebagai calon kedua puluh empat untuk temu duga pada tujuh Mei."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah dipilih sebagai calon ke-31 untuk temu duga pada 27 Disember."}, {"role": "assistant", "content": "Anda telah dipilih sebagai calon ketiga puluh satu untuk temu duga pada dua puluh tujuh Disember."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari 7 malam hingga 1:05 pada 21 Ogos."}, {"role": "assistant", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari tujuh malam hingga satu kosong lima pada dua puluh satu Ogos."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda adalah pesakit ke-8 yang dijadualkan untuk sesi konsultasi hari ini."}, {"role": "assistant", "content": "Anda adalah pesakit kelapan yang dijadualkan untuk sesi konsultasi hari ini."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah terpilih untuk menerima rebat sebanyak 66% bagi langganan tahun 2002."}, {"role": "assistant", "content": "Anda telah terpilih untuk menerima rebat sebanyak enam puluh enam peratus bagi langganan tahun dua ribu dua."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah terpilih untuk menerima rebat sebanyak 49% bagi langganan tahun 1964."}, {"role": "assistant", "content": "Anda telah terpilih untuk menerima rebat sebanyak empat puluh sembilan peratus bagi langganan tahun seribu sembilan ratus enam puluh empat."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Barangan anda telah sampai di pusat agihan ke-11 sebelum dihantar ke rumah."}, {"role": "assistant", "content": "Barangan anda telah sampai di pusat agihan kesebelas sebelum dihantar ke rumah."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah terpilih untuk menerima rebat sebanyak 17.1% bagi langganan tahun 2005."}, {"role": "assistant", "content": "Anda telah terpilih untuk menerima rebat sebanyak tujuh belas perpuluhan satu peratus bagi langganan tahun dua ribu lima."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah dipilih sebagai calon ke-21 untuk temu duga pada 23 Disember."}, {"role": "assistant", "content": "Anda telah dipilih sebagai calon kedua puluh satu untuk temu duga pada dua puluh tiga Disember."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah terpilih untuk menerima rebat sebanyak 3% bagi langganan tahun 1961."}, {"role": "assistant", "content": "Anda telah terpilih untuk menerima rebat sebanyak tiga peratus bagi langganan tahun seribu sembilan ratus enam puluh satu."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah dipilih sebagai calon ke-21 untuk temu duga pada 24/10/2014."}, {"role": "assistant", "content": "Anda telah dipilih sebagai calon kedua puluh satu untuk temu duga pada dua puluh empat Oktober dua ribu empat belas."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah dipilih sebagai calon ke-15 untuk temu duga pada 23/4/2025."}, {"role": "assistant", "content": "Anda telah dipilih sebagai calon kelima belas untuk temu duga pada dua puluh tiga April dua ribu dua puluh lima."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah dipilih sebagai calon ke-11 untuk temu duga pada 15/9/2017."}, {"role": "assistant", "content": "Anda telah dipilih sebagai calon kesebelas untuk temu duga pada lima belas September dua ribu tujuh belas."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah dipilih sebagai calon ke-3 untuk temu duga pada 8 September."}, {"role": "assistant", "content": "Anda telah dipilih sebagai calon ketiga untuk temu duga pada lapan September."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah terpilih untuk menerima rebat sebanyak 7.9% bagi langganan tahun 2028."}, {"role": "assistant", "content": "Anda telah terpilih untuk menerima rebat sebanyak tujuh perpuluhan sembilan peratus bagi langganan tahun dua ribu dua puluh lapan."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Barangan anda telah sampai di pusat agihan ke-25 sebelum dihantar ke rumah."}, {"role": "assistant", "content": "Barangan anda telah sampai di pusat agihan kedua puluh lima sebelum dihantar ke rumah."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah terpilih untuk menerima rebat sebanyak 55% bagi langganan tahun 1954."}, {"role": "assistant", "content": "Anda telah terpilih untuk menerima rebat sebanyak lima puluh lima peratus bagi langganan tahun seribu sembilan ratus lima puluh empat."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda adalah pesakit ke-7 yang dijadualkan untuk sesi konsultasi hari ini."}, {"role": "assistant", "content": "Anda adalah pesakit ketujuh yang dijadualkan untuk sesi konsultasi hari ini."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah terpilih untuk menerima rebat sebanyak 21% bagi langganan tahun 2008."}, {"role": "assistant", "content": "Anda telah terpilih untuk menerima rebat sebanyak dua puluh satu peratus bagi langganan tahun dua ribu lapan."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah terpilih untuk menerima rebat sebanyak 96% bagi langganan tahun 1975."}, {"role": "assistant", "content": "Anda telah terpilih untuk menerima rebat sebanyak sembilan puluh enam peratus bagi langganan tahun seribu sembilan ratus tujuh puluh lima."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda adalah pesakit ke-11 yang dijadualkan untuk sesi konsultasi hari ini."}, {"role": "assistant", "content": "Anda adalah pesakit kesebelas yang dijadualkan untuk sesi konsultasi hari ini."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah dipilih sebagai calon ke-3 untuk temu duga pada 22-08-2025."}, {"role": "assistant", "content": "Anda telah dipilih sebagai calon ketiga untuk temu duga pada dua puluh dua Ogos dua ribu dua puluh lima."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah terpilih untuk menerima rebat sebanyak 76% bagi langganan tahun 1968."}, {"role": "assistant", "content": "Anda telah terpilih untuk menerima rebat sebanyak tujuh puluh enam peratus bagi langganan tahun seribu sembilan ratus enam puluh lapan."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda adalah pesakit ke-16 yang dijadualkan untuk sesi konsultasi hari ini."}, {"role": "assistant", "content": "Anda adalah pesakit keenam belas yang dijadualkan untuk sesi konsultasi hari ini."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari 3:40 petang hingga 8.05 pagi pada 16-05-2002."}, {"role": "assistant", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari tiga empat puluh petang hingga lapan kosong lima pagi pada enam belas Mei dua ribu dua."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah terpilih untuk menerima rebat sebanyak 50% bagi langganan tahun 1995."}, {"role": "assistant", "content": "Anda telah terpilih untuk menerima rebat sebanyak lima puluh peratus bagi langganan tahun seribu sembilan ratus sembilan puluh lima."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah terpilih untuk menerima rebat sebanyak 84% bagi langganan tahun 1982."}, {"role": "assistant", "content": "Anda telah terpilih untuk menerima rebat sebanyak lapan puluh empat peratus bagi langganan tahun seribu sembilan ratus lapan puluh dua."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah dipilih sebagai calon ke-26 untuk temu duga pada 11 November."}, {"role": "assistant", "content": "Anda telah dipilih sebagai calon kedua puluh enam untuk temu duga pada sebelas November."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari 7:10 hingga 4.50 malam pada 22 Mac 2028."}, {"role": "assistant", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari tujuh sepuluh hingga empat lima puluh malam pada dua puluh dua Mac dua ribu dua puluh lapan."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah terpilih untuk menerima rebat sebanyak 85% bagi langganan tahun 1950."}, {"role": "assistant", "content": "Anda telah terpilih untuk menerima rebat sebanyak lapan puluh lima peratus bagi langganan tahun seribu sembilan ratus lima puluh."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah terpilih untuk menerima rebat sebanyak 78% bagi langganan tahun 1975."}, {"role": "assistant", "content": "Anda telah terpilih untuk menerima rebat sebanyak tujuh puluh lapan peratus bagi langganan tahun seribu sembilan ratus tujuh puluh lima."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari 2.40 petang hingga 11 pagi pada 3 Mei."}, {"role": "assistant", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari dua empat puluh petang hingga sebelas pagi pada tiga Mei."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah terpilih untuk menerima rebat sebanyak 68% bagi langganan tahun 2009."}, {"role": "assistant", "content": "Anda telah terpilih untuk menerima rebat sebanyak enam puluh lapan peratus bagi langganan tahun dua ribu sembilan."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah terpilih untuk menerima rebat sebanyak 1% bagi langganan tahun 2026."}, {"role": "assistant", "content": "Anda telah terpilih untuk menerima rebat sebanyak satu peratus bagi langganan tahun dua ribu dua puluh enam."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda adalah pesakit ke-20 yang dijadualkan untuk sesi konsultasi hari ini."}, {"role": "assistant", "content": "Anda adalah pesakit kedua puluh yang dijadualkan untuk sesi konsultasi hari ini."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Barangan anda telah sampai di pusat agihan ke-22 sebelum dihantar ke rumah."}, {"role": "assistant", "content": "Barangan anda telah sampai di pusat agihan kedua puluh dua sebelum dihantar ke rumah."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Barangan anda telah sampai di pusat agihan ke-17 sebelum dihantar ke rumah."}, {"role": "assistant", "content": "Barangan anda telah sampai di pusat agihan ketujuh belas sebelum dihantar ke rumah."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda adalah pesakit ke-10 yang dijadualkan untuk sesi konsultasi hari ini."}, {"role": "assistant", "content": "Anda adalah pesakit kesepuluh yang dijadualkan untuk sesi konsultasi hari ini."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Barangan anda telah sampai di pusat agihan ke-18 sebelum dihantar ke rumah."}, {"role": "assistant", "content": "Barangan anda telah sampai di pusat agihan kelapan belas sebelum dihantar ke rumah."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah dipilih sebagai calon ke-27 untuk temu duga pada 15 Julai."}, {"role": "assistant", "content": "Anda telah dipilih sebagai calon kedua puluh tujuh untuk temu duga pada lima belas Julai."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari 7:10 hingga 9 pagi pada 13/10/2006."}, {"role": "assistant", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari tujuh sepuluh hingga sembilan pagi pada tiga belas Oktober dua ribu enam."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari 5.00 pagi hingga 2:00 pada 18 Disember."}, {"role": "assistant", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari lima pagi hingga dua pada lapan belas Disember."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari 3:00 malam hingga 10:05 pagi pada 13-01-2023."}, {"role": "assistant", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari tiga malam hingga sepuluh kosong lima pagi pada tiga belas Januari dua ribu dua puluh tiga."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Barangan anda telah sampai di pusat agihan ke-10 sebelum dihantar ke rumah."}, {"role": "assistant", "content": "Barangan anda telah sampai di pusat agihan kesepuluh sebelum dihantar ke rumah."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah terpilih untuk menerima rebat sebanyak 68% bagi langganan tahun 2014."}, {"role": "assistant", "content": "Anda telah terpilih untuk menerima rebat sebanyak enam puluh lapan peratus bagi langganan tahun dua ribu empat belas."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah terpilih untuk menerima rebat sebanyak 60% bagi langganan tahun 2027."}, {"role": "assistant", "content": "Anda telah terpilih untuk menerima rebat sebanyak enam puluh peratus bagi langganan tahun dua ribu dua puluh tujuh."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda adalah pesakit ke-5 yang dijadualkan untuk sesi konsultasi hari ini."}, {"role": "assistant", "content": "Anda adalah pesakit kelima yang dijadualkan untuk sesi konsultasi hari ini."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari 7:00 pagi hingga 2:30 petang pada 6 Ogos."}, {"role": "assistant", "content": "Tempoh penyelenggaraan sistem akan berlangsung dari tujuh pagi hingga dua tiga puluh petang pada enam Ogos."}], "lang": "ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 760557, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah tujuh enam nol lima lima tujuh, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul 21-26."}, {"role": "assistant", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul dua puluh satu sampai dua puluh enam."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul 12 - 13."}, {"role": "assistant", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul dua belas sampai tiga belas."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paket internet ini berlaku selama 1-4 hari sejak tanggal aktivasi."}, {"role": "assistant", "content": "Paket internet ini berlaku selama satu sampai empat hari sejak tanggal aktivasi."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 449.395 byte data dari total kuota 2l Anda."}, {"role": "assistant", "content": "Anda telah menggunakan empat ratus empat puluh sembilan ribu tiga ratus sembilan puluh lima byte data dari total kuota dua liter Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paket internet ini berlaku selama 3-6 hari sejak tanggal aktivasi."}, {"role": "assistant", "content": "Paket internet ini berlaku selama tiga sampai enam hari sejak tanggal aktivasi."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada 15 Oktober 2023 pukul 18.15."}, {"role": "assistant", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada lima belas Oktober dua ribu dua puluh tiga pukul delapan belas lima belas."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paket internet ini berlaku selama 11–21 hari sejak tanggal aktivasi."}, {"role": "assistant", "content": "Paket internet ini berlaku selama sebelas sampai dua puluh satu hari sejak tanggal aktivasi."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 673556, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah enam tujuh tiga lima lima enam, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 9861, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah sembilan delapan enam satu, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 2.529 byte data dari total kuota 5 GB Anda."}, {"role": "assistant", "content": "Anda telah menggunakan dua ribu lima ratus dua puluh sembilan byte data dari total kuota lima gigabyte Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 52176, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah lima dua satu tujuh enam, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada 17/07/2028 pukul 16:40."}, {"role": "assistant", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada tujuh belas Juli dua ribu dua puluh delapan pukul enam belas empat puluh."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paket internet ini berlaku selama 11-19 hari sejak tanggal aktivasi."}, {"role": "assistant", "content": "Paket internet ini berlaku selama sebelas sampai sembilan belas hari sejak tanggal aktivasi."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 71833, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah tujuh satu delapan tiga tiga, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paket internet ini berlaku selama 30-36 hari sejak tanggal aktivasi."}, {"role": "assistant", "content": "Paket internet ini berlaku selama tiga puluh sampai tiga puluh enam hari sejak tanggal aktivasi."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 34728, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah tiga empat tujuh dua delapan, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 705.579 byte data dari total kuota 256 ml Anda."}, {"role": "assistant", "content": "Anda telah menggunakan tujuh ratus lima ribu lima ratus tujuh puluh sembilan byte data dari total kuota dua ratus lima puluh enam mililiter Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 17.179 byte data dari total kuota 1,6km Anda."}, {"role": "assistant", "content": "Anda telah menggunakan tujuh belas ribu seratus tujuh puluh sembilan byte data dari total kuota satu koma enam kilometer Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul 9-18."}, {"role": "assistant", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul sembilan sampai delapan belas."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paket internet ini berlaku selama 5–6 hari sejak tanggal aktivasi."}, {"role": "assistant", "content": "Paket internet ini berlaku selama lima sampai enam hari sejak tanggal aktivasi."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paket internet ini berlaku selama 8–14 hari sejak tanggal aktivasi."}, {"role": "assistant", "content": "Paket internet ini berlaku selama delapan sampai empat belas hari sejak tanggal aktivasi."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul 22 - 28."}, {"role": "assistant", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul dua puluh dua sampai dua puluh delapan."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 89.073 byte data dari total kuota 5,3kg Anda."}, {"role": "assistant", "content": "Anda telah menggunakan delapan puluh sembilan ribu tujuh puluh tiga byte data dari total kuota lima koma tiga kilogram Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 30.117 byte data dari total kuota 1,3km Anda."}, {"role": "assistant", "content": "Anda telah menggunakan tiga puluh ribu seratus tujuh belas byte data dari total kuota satu koma tiga kilometer Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paket internet ini berlaku selama 21–28 hari sejak tanggal aktivasi."}, {"role": "assistant", "content": "Paket internet ini berlaku selama dua puluh satu sampai dua puluh delapan hari sejak tanggal aktivasi."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 98.532 byte data dari total kuota 9 g Anda."}, {"role": "assistant", "content": "Anda telah menggunakan sembilan puluh delapan ribu lima ratus tiga puluh dua byte data dari total kuota sembilan gram Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 471005, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah empat tujuh satu nol nol lima, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul 4–11."}, {"role": "assistant", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul empat sampai sebelas."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paket internet ini berlaku selama 25 - 28 hari sejak tanggal aktivasi."}, {"role": "assistant", "content": "Paket internet ini berlaku selama dua puluh lima sampai dua puluh delapan hari sejak tanggal aktivasi."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paket internet ini berlaku selama 24 - 33 hari sejak tanggal aktivasi."}, {"role": "assistant", "content": "Paket internet ini berlaku selama dua puluh empat sampai tiga puluh tiga hari sejak tanggal aktivasi."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 060406, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah nol enam nol empat nol enam, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 2562, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah dua lima enam dua, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 61.571 byte data dari total kuota 1km Anda."}, {"role": "assistant", "content": "Anda telah menggunakan enam puluh satu ribu lima ratus tujuh puluh satu byte data dari total kuota satu kilometer Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paket internet ini berlaku selama 7-13 hari sejak tanggal aktivasi."}, {"role": "assistant", "content": "Paket internet ini berlaku selama tujuh sampai tiga belas hari sejak tanggal aktivasi."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paket internet ini berlaku selama 8–16 hari sejak tanggal aktivasi."}, {"role": "assistant", "content": "Paket internet ini berlaku selama delapan sampai enam belas hari sejak tanggal aktivasi."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 54.665.460 byte data dari total kuota 320,8kg Anda."}, {"role": "assistant", "content": "Anda telah menggunakan lima puluh empat juta enam ratus enam puluh lima ribu empat ratus enam puluh byte data dari total kuota tiga ratus dua puluh koma delapan kilogram Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul 10-17."}, {"role": "assistant", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul sepuluh sampai tujuh belas."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 730306, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah tujuh tiga nol tiga nol enam, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 766575, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah tujuh enam enam lima tujuh lima, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paket internet ini berlaku selama 4-5 hari sejak tanggal aktivasi."}, {"role": "assistant", "content": "Paket internet ini berlaku selama empat sampai lima hari sejak tanggal aktivasi."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 77.536 byte data dari total kuota 34,9 GB Anda."}, {"role": "assistant", "content": "Anda telah menggunakan tujuh puluh tujuh ribu lima ratus tiga puluh enam byte data dari total kuota tiga puluh empat koma sembilan gigabyte Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada 03-12-2000 pukul 10."}, {"role": "assistant", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada tiga Desember dua ribu pukul sepuluh."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 32.987.947 byte data dari total kuota 7,1GB Anda."}, {"role": "assistant", "content": "Anda telah menggunakan tiga puluh dua juta sembilan ratus delapan puluh tujuh ribu sembilan ratus empat puluh tujuh byte data dari total kuota tujuh koma satu gigabyte Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul 19–28."}, {"role": "assistant", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul sembilan belas sampai dua puluh delapan."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul 26-27."}, {"role": "assistant", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul dua puluh enam sampai dua puluh tujuh."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada 4 Maret 2025 pukul 18.30."}, {"role": "assistant", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada empat Maret dua ribu dua puluh lima pukul delapan belas tiga puluh."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul 5 - 13."}, {"role": "assistant", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul lima sampai tiga belas."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul 4 - 11."}, {"role": "assistant", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul empat sampai sebelas."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 41.911 byte data dari total kuota 1 L Anda."}, {"role": "assistant", "content": "Anda telah menggunakan empat puluh satu ribu sembilan ratus sebelas byte data dari total kuota satu liter Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul 13-19."}, {"role": "assistant", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul tiga belas sampai sembilan belas."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 73.382 byte data dari total kuota 42,5 GB Anda."}, {"role": "assistant", "content": "Anda telah menggunakan tujuh puluh tiga ribu tiga ratus delapan puluh dua byte data dari total kuota empat puluh dua koma lima gigabyte Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 62.645 byte data dari total kuota 439,1 kg Anda."}, {"role": "assistant", "content": "Anda telah menggunakan enam puluh dua ribu enam ratus empat puluh lima byte data dari total kuota empat ratus tiga puluh sembilan koma satu kilogram Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 952.542 byte data dari total kuota 9 kg Anda."}, {"role": "assistant", "content": "Anda telah menggunakan sembilan ratus lima puluh dua ribu lima ratus empat puluh dua byte data dari total kuota sembilan kilogram Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 111.142 byte data dari total kuota 3 ml Anda."}, {"role": "assistant", "content": "Anda telah menggunakan seratus sebelas ribu seratus empat puluh dua byte data dari total kuota tiga mililiter Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paket internet ini berlaku selama 8-10 hari sejak tanggal aktivasi."}, {"role": "assistant", "content": "Paket internet ini berlaku selama delapan sampai sepuluh hari sejak tanggal aktivasi."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 408.292 byte data dari total kuota 3ml Anda."}, {"role": "assistant", "content": "Anda telah menggunakan empat ratus delapan ribu dua ratus sembilan puluh dua byte data dari total kuota tiga mililiter Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada 16-12-2027 pukul 14:50."}, {"role": "assistant", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada enam belas Desember dua ribu dua puluh tujuh pukul empat belas lima puluh."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada 2/02/2025 pukul 12.20."}, {"role": "assistant", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada dua Februari dua ribu dua puluh lima pukul dua belas dua puluh."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 397305, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah tiga sembilan tujuh tiga nol lima, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 261445, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah dua enam satu empat empat lima, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul 15-24."}, {"role": "assistant", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul lima belas sampai dua puluh empat."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paket internet ini berlaku selama 12 - 16 hari sejak tanggal aktivasi."}, {"role": "assistant", "content": "Paket internet ini berlaku selama dua belas sampai enam belas hari sejak tanggal aktivasi."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paket internet ini berlaku selama 4 - 8 hari sejak tanggal aktivasi."}, {"role": "assistant", "content": "Paket internet ini berlaku selama empat sampai delapan hari sejak tanggal aktivasi."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paket internet ini berlaku selama 10 - 15 hari sejak tanggal aktivasi."}, {"role": "assistant", "content": "Paket internet ini berlaku selama sepuluh sampai lima belas hari sejak tanggal aktivasi."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 714.241 byte data dari total kuota 376 gr Anda."}, {"role": "assistant", "content": "Anda telah menggunakan tujuh ratus empat belas ribu dua ratus empat puluh satu byte data dari total kuota tiga ratus tujuh puluh enam gram Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada 23-09-2013 pukul 18.40."}, {"role": "assistant", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada dua puluh tiga September dua ribu tiga belas pukul delapan belas empat puluh."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 44.322 byte data dari total kuota 239 km Anda."}, {"role": "assistant", "content": "Anda telah menggunakan empat puluh empat ribu tiga ratus dua puluh dua byte data dari total kuota dua ratus tiga puluh sembilan kilometer Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul 24-33."}, {"role": "assistant", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul dua puluh empat sampai tiga puluh tiga."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 787.074 byte data dari total kuota 1,1kg Anda."}, {"role": "assistant", "content": "Anda telah menggunakan tujuh ratus delapan puluh tujuh ribu tujuh puluh empat byte data dari total kuota satu koma satu kilogram Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul 25 - 33."}, {"role": "assistant", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul dua puluh lima sampai tiga puluh tiga."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul 26–32."}, {"role": "assistant", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul dua puluh enam sampai tiga puluh dua."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 5026 byte data dari total kuota 1l Anda."}, {"role": "assistant", "content": "Anda telah menggunakan lima ribu dua puluh enam byte data dari total kuota satu liter Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada 16 November 2008 pukul 8.00."}, {"role": "assistant", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada enam belas November dua ribu delapan pukul delapan."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 363157, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah tiga enam tiga satu lima tujuh, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul 4 - 9."}, {"role": "assistant", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul empat sampai sembilan."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 7067, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah tujuh nol enam tujuh, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 35538, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah tiga lima lima tiga delapan, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 63993, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah enam tiga sembilan sembilan tiga, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 2105 byte data dari total kuota 1km Anda."}, {"role": "assistant", "content": "Anda telah menggunakan dua ribu seratus lima byte data dari total kuota satu kilometer Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 9259, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah sembilan dua lima sembilan, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul 12-20."}, {"role": "assistant", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul dua belas sampai dua puluh."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paket internet ini berlaku selama 25-35 hari sejak tanggal aktivasi."}, {"role": "assistant", "content": "Paket internet ini berlaku selama dua puluh lima sampai tiga puluh lima hari sejak tanggal aktivasi."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paket internet ini berlaku selama 13-14 hari sejak tanggal aktivasi."}, {"role": "assistant", "content": "Paket internet ini berlaku selama tiga belas sampai empat belas hari sejak tanggal aktivasi."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 33836, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah tiga tiga delapan tiga enam, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 12855, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah satu dua delapan lima lima, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paket internet ini berlaku selama 16-24 hari sejak tanggal aktivasi."}, {"role": "assistant", "content": "Paket internet ini berlaku selama enam belas sampai dua puluh empat hari sejak tanggal aktivasi."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul 22-26."}, {"role": "assistant", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul dua puluh dua sampai dua puluh enam."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada 15 Juni 2016 pukul 3:00."}, {"role": "assistant", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada lima belas Juni dua ribu enam belas pukul tiga."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada 23-05-2029 pukul 8.15."}, {"role": "assistant", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada dua puluh tiga Mei dua ribu dua puluh sembilan pukul delapan lima belas."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul 28-33."}, {"role": "assistant", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul dua puluh delapan sampai tiga puluh tiga."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 5.079 byte data dari total kuota 1 °C Anda."}, {"role": "assistant", "content": "Anda telah menggunakan lima ribu tujuh puluh sembilan byte data dari total kuota satu derajat Celsius Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 627727, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah enam dua tujuh tujuh dua tujuh, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 50.509 byte data dari total kuota 7GB Anda."}, {"role": "assistant", "content": "Anda telah menggunakan lima puluh ribu lima ratus sembilan byte data dari total kuota tujuh gigabyte Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada 20 April pukul 19."}, {"role": "assistant", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada dua puluh April pukul sembilan belas."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 80.993 byte data dari total kuota 1 cm Anda."}, {"role": "assistant", "content": "Anda telah menggunakan delapan puluh ribu sembilan ratus sembilan puluh tiga byte data dari total kuota satu sentimeter Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada 19-10-2010 pukul 7."}, {"role": "assistant", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada sembilan belas Oktober dua ribu sepuluh pukul tujuh."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 16.588 byte data dari total kuota 113,6kg Anda."}, {"role": "assistant", "content": "Anda telah menggunakan enam belas ribu lima ratus delapan puluh delapan byte data dari total kuota seratus tiga belas koma enam kilogram Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 4.591 byte data dari total kuota 453,8 ml Anda."}, {"role": "assistant", "content": "Anda telah menggunakan empat ribu lima ratus sembilan puluh satu byte data dari total kuota empat ratus lima puluh tiga koma delapan mililiter Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada 14/05/2006 pukul 4:05."}, {"role": "assistant", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada empat belas Mei dua ribu enam pukul empat nol lima."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 56090, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah lima enam nol sembilan nol, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paket internet ini berlaku selama 3-9 hari sejak tanggal aktivasi."}, {"role": "assistant", "content": "Paket internet ini berlaku selama tiga sampai sembilan hari sejak tanggal aktivasi."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 0771, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah nol tujuh tujuh satu, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 32.810.520 byte data dari total kuota 1,7m Anda."}, {"role": "assistant", "content": "Anda telah menggunakan tiga puluh dua juta delapan ratus sepuluh ribu lima ratus dua puluh byte data dari total kuota satu koma tujuh meter Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paket internet ini berlaku selama 10 - 20 hari sejak tanggal aktivasi."}, {"role": "assistant", "content": "Paket internet ini berlaku selama sepuluh sampai dua puluh hari sejak tanggal aktivasi."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paket internet ini berlaku selama 26 - 28 hari sejak tanggal aktivasi."}, {"role": "assistant", "content": "Paket internet ini berlaku selama dua puluh enam sampai dua puluh delapan hari sejak tanggal aktivasi."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 86789, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah delapan enam tujuh delapan sembilan, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada 9/07/2031 pukul 2:45."}, {"role": "assistant", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada sembilan Juli dua ribu tiga puluh satu pukul dua empat puluh lima."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 42.562.709 byte data dari total kuota 4 g Anda."}, {"role": "assistant", "content": "Anda telah menggunakan empat puluh dua juta lima ratus enam puluh dua ribu tujuh ratus sembilan byte data dari total kuota empat gram Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 035461, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah nol tiga lima empat enam satu, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada 23-02-2012 pukul 21.05."}, {"role": "assistant", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada dua puluh tiga Februari dua ribu dua belas pukul dua puluh satu nol lima."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 5498, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah lima empat sembilan delapan, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 434.738 byte data dari total kuota 352,3kg Anda."}, {"role": "assistant", "content": "Anda telah menggunakan empat ratus tiga puluh empat ribu tujuh ratus tiga puluh delapan byte data dari total kuota tiga ratus lima puluh dua koma tiga kilogram Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paket internet ini berlaku selama 9-10 hari sejak tanggal aktivasi."}, {"role": "assistant", "content": "Paket internet ini berlaku selama sembilan sampai sepuluh hari sejak tanggal aktivasi."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul 4-6."}, {"role": "assistant", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul empat sampai enam."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 4958, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah empat sembilan lima delapan, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 90605, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah sembilan nol enam nol lima, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada 8 Desember 2017 pukul 15:00."}, {"role": "assistant", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada delapan Desember dua ribu tujuh belas pukul lima belas."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 5829, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah lima delapan dua sembilan, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paket internet ini berlaku selama 16-25 hari sejak tanggal aktivasi."}, {"role": "assistant", "content": "Paket internet ini berlaku selama enam belas sampai dua puluh lima hari sejak tanggal aktivasi."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 724642, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah tujuh dua empat enam empat dua, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 157.394 byte data dari total kuota 2,4km Anda."}, {"role": "assistant", "content": "Anda telah menggunakan seratus lima puluh tujuh ribu tiga ratus sembilan puluh empat byte data dari total kuota dua koma empat kilometer Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 768789, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah tujuh enam delapan tujuh delapan sembilan, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 6026 byte data dari total kuota 4 cm Anda."}, {"role": "assistant", "content": "Anda telah menggunakan enam ribu dua puluh enam byte data dari total kuota empat sentimeter Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada 2 Agustus pukul 18.10."}, {"role": "assistant", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada dua Agustus pukul delapan belas sepuluh."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 4.029 byte data dari total kuota 8,6 km Anda."}, {"role": "assistant", "content": "Anda telah menggunakan empat ribu dua puluh sembilan byte data dari total kuota delapan koma enam kilometer Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul 30-33."}, {"role": "assistant", "content": "Layanan pengaduan kami tersedia setiap hari Senin sampai Jumat pukul tiga puluh sampai tiga puluh tiga."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Anda telah menggunakan 429.143 byte data dari total kuota 5cm Anda."}, {"role": "assistant", "content": "Anda telah menggunakan empat ratus dua puluh sembilan ribu seratus empat puluh tiga byte data dari total kuota lima sentimeter Anda."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada 8 Januari pukul 10:00."}, {"role": "assistant", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada delapan Januari pukul sepuluh."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada 10 Desember pukul 20.30."}, {"role": "assistant", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada sepuluh Desember pukul dua puluh tiga puluh."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 8850, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah delapan delapan lima nol, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 0992, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah nol sembilan sembilan dua, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paket internet ini berlaku selama 24–29 hari sejak tanggal aktivasi."}, {"role": "assistant", "content": "Paket internet ini berlaku selama dua puluh empat sampai dua puluh sembilan hari sejak tanggal aktivasi."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada 18/11/2027 pukul 19:30."}, {"role": "assistant", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada delapan belas November dua ribu dua puluh tujuh pukul sembilan belas tiga puluh."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 02866, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah nol dua delapan enam enam, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada 12 September pukul 5:20."}, {"role": "assistant", "content": "Halo, jadwal konsultasi Anda dengan dokter adalah pada dua belas September pukul lima dua puluh."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 771294, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah tujuh tujuh satu dua sembilan empat, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kode verifikasi pendaftaran Anda adalah 645021, mohon jangan berikan kepada siapa pun."}, {"role": "assistant", "content": "Kode verifikasi pendaftaran Anda adalah enam empat lima nol dua satu, mohon jangan berikan kepada siapa pun."}], "lang": "id", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "快递员将在下午7:30左右到达，请在27-35之间留意接收。"}, {"role": "assistant", "content": "快递员将在下午七点三十分左右到达，请在二十七到三十五之间留意接收。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "快递员将在上午7点30分左右到达，请在21–25之间留意接收。"}, {"role": "assistant", "content": "快递员将在上午七点三十分左右到达，请在二十一到二十五之间留意接收。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "由于人数较多，预计等待时间为62.97小时。"}, {"role": "assistant", "content": "由于人数较多，预计等待时间为六十二点九七小时。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您排在第1位，还需等待约50分钟。"}, {"role": "assistant", "content": "您排在第一位，还需等待约五十分钟。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "由于人数较多，预计等待时间为78.2小时。"}, {"role": "assistant", "content": "由于人数较多，预计等待时间为七十八点二小时。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请在14 - 21个工作日内等待审核结果。"}, {"role": "assistant", "content": "请在十四到二十一个工作日内等待审核结果。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您好，您的账户余额已低于RM78,192.56，请及时充值。"}, {"role": "assistant", "content": "您好，您的账户余额已低于七万八千一百九十二令吉五十六仙，请及时充值。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "快递员将在上午5:50左右到达，请在28-38之间留意接收。"}, {"role": "assistant", "content": "快递员将在上午五点五十分左右到达，请在二十八到三十八之间留意接收。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您排在第23位，还需等待约41分钟。"}, {"role": "assistant", "content": "您排在第二十三位，还需等待约四十一分钟。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请在27-37个工作日内等待审核结果。"}, {"role": "assistant", "content": "请在二十七到三十七个工作日内等待审核结果。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请在13 - 18个工作日内等待审核结果。"}, {"role": "assistant", "content": "请在十三到十八个工作日内等待审核结果。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "快递员将在2点左右到达，请在10–19之间留意接收。"}, {"role": "assistant", "content": "快递员将在两点左右到达，请在十到十九之间留意接收。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请在10–11个工作日内等待审核结果。"}, {"role": "assistant", "content": "请在十到十一个工作日内等待审核结果。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您好，您的账户余额已低于RM455.96，请及时充值。"}, {"role": "assistant", "content": "您好，您的账户余额已低于四百五十五令吉九十六仙，请及时充值。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您排在第6位，还需等待约67分钟。"}, {"role": "assistant", "content": "您排在第六位，还需等待约六十七分钟。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您好，您的账户余额已低于RM83,057.90，请及时充值。"}, {"role": "assistant", "content": "您好，您的账户余额已低于八万三千零五十七令吉九十仙，请及时充值。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您排在第8位，还需等待约35分钟。"}, {"role": "assistant", "content": "您排在第八位，还需等待约三十五分钟。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您好，您的账户余额已低于RM 375.69，请及时充值。"}, {"role": "assistant", "content": "您好，您的账户余额已低于三百七十五令吉六十九仙，请及时充值。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您排在第6位，还需等待约21分钟。"}, {"role": "assistant", "content": "您排在第六位，还需等待约二十一分钟。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您好，您的账户余额已低于RM937，请及时充值。"}, {"role": "assistant", "content": "您好，您的账户余额已低于九百三十七令吉，请及时充值。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "快递员将在下午8点50分左右到达，请在25 - 29之间留意接收。"}, {"role": "assistant", "content": "快递员将在下午八点五十分左右到达，请在二十五到二十九之间留意接收。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您好，您的账户余额已低于RM20,301.90，请及时充值。"}, {"role": "assistant", "content": "您好，您的账户余额已低于二万零三百零一令吉九十仙，请及时充值。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "由于人数较多，预计等待时间为36.5小时。"}, {"role": "assistant", "content": "由于人数较多，预计等待时间为三十六点五小时。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您好，您的账户余额已低于RM628.50，请及时充值。"}, {"role": "assistant", "content": "您好，您的账户余额已低于六百二十八令吉五十仙，请及时充值。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您排在第11位，还需等待约79分钟。"}, {"role": "assistant", "content": "您排在第十一位，还需等待约七十九分钟。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您好，您的账户余额已低于RM42,746.50，请及时充值。"}, {"role": "assistant", "content": "您好，您的账户余额已低于四万二千七百四十六令吉五十仙，请及时充值。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请在30–40个工作日内等待审核结果。"}, {"role": "assistant", "content": "请在三十到四十个工作日内等待审核结果。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "由于人数较多，预计等待时间为38.9小时。"}, {"role": "assistant", "content": "由于人数较多，预计等待时间为三十八点九小时。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "由于人数较多，预计等待时间为4.3小时。"}, {"role": "assistant", "content": "由于人数较多，预计等待时间为四点三小时。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您排在第31位，还需等待约6分钟。"}, {"role": "assistant", "content": "您排在第三十一位，还需等待约六分钟。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "由于人数较多，预计等待时间为34.91小时。"}, {"role": "assistant", "content": "由于人数较多，预计等待时间为三十四点九一小时。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "由于人数较多，预计等待时间为3.44小时。"}, {"role": "assistant", "content": "由于人数较多，预计等待时间为三点四四小时。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请在29 - 39个工作日内等待审核结果。"}, {"role": "assistant", "content": "请在二十九到三十九个工作日内等待审核结果。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "由于人数较多，预计等待时间为4.41小时。"}, {"role": "assistant", "content": "由于人数较多，预计等待时间为四点四一小时。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "快递员将在下午9点45分左右到达，请在4-13之间留意接收。"}, {"role": "assistant", "content": "快递员将在下午九点四十五分左右到达，请在四到十三之间留意接收。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您好，您的账户余额已低于RM51,886，请及时充值。"}, {"role": "assistant", "content": "您好，您的账户余额已低于五万一千八百八十六令吉，请及时充值。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您好，您的账户余额已低于RM89,260，请及时充值。"}, {"role": "assistant", "content": "您好，您的账户余额已低于八万九千二百六十令吉，请及时充值。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您排在第30位，还需等待约2分钟。"}, {"role": "assistant", "content": "您排在第三十位，还需等待约二分钟。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "快递员将在下午2:20左右到达，请在11 - 19之间留意接收。"}, {"role": "assistant", "content": "快递员将在下午两点二十分左右到达，请在十一到十九之间留意接收。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您排在第31位，还需等待约719分钟。"}, {"role": "assistant", "content": "您排在第三十一位，还需等待约七百一十九分钟。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请在30 - 36个工作日内等待审核结果。"}, {"role": "assistant", "content": "请在三十到三十六个工作日内等待审核结果。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "由于人数较多，预计等待时间为82.1小时。"}, {"role": "assistant", "content": "由于人数较多，预计等待时间为八十二点一小时。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "快递员将在19:50左右到达，请在4–12之间留意接收。"}, {"role": "assistant", "content": "快递员将在十九点五十分左右到达，请在四到十二之间留意接收。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请在26-34个工作日内等待审核结果。"}, {"role": "assistant", "content": "请在二十六到三十四个工作日内等待审核结果。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请在7–11个工作日内等待审核结果。"}, {"role": "assistant", "content": "请在七到十一个工作日内等待审核结果。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "快递员将在5点左右到达，请在22-31之间留意接收。"}, {"role": "assistant", "content": "快递员将在五点左右到达，请在二十二到三十一之间留意接收。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "由于人数较多，预计等待时间为53.75小时。"}, {"role": "assistant", "content": "由于人数较多，预计等待时间为五十三点七五小时。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "由于人数较多，预计等待时间为31.22小时。"}, {"role": "assistant", "content": "由于人数较多，预计等待时间为三十一点二二小时。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您好，您的账户余额已低于RM40，请及时充值。"}, {"role": "assistant", "content": "您好，您的账户余额已低于四十令吉，请及时充值。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "由于人数较多，预计等待时间为20.40小时。"}, {"role": "assistant", "content": "由于人数较多，预计等待时间为二十点四零小时。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您好，您的账户余额已低于RM201，请及时充值。"}, {"role": "assistant", "content": "您好，您的账户余额已低于二百零一令吉，请及时充值。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您排在第25位，还需等待约903分钟。"}, {"role": "assistant", "content": "您排在第二十五位，还需等待约九百零三分钟。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请在26-27个工作日内等待审核结果。"}, {"role": "assistant", "content": "请在二十六到二十七个工作日内等待审核结果。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您好，您的账户余额已低于RM87.33，请及时充值。"}, {"role": "assistant", "content": "您好，您的账户余额已低于八十七令吉三十三仙，请及时充值。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "快递员将在6:20左右到达，请在14-19之间留意接收。"}, {"role": "assistant", "content": "快递员将在六点二十分左右到达，请在十四到十九之间留意接收。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您好，您的账户余额已低于RM12,766.50，请及时充值。"}, {"role": "assistant", "content": "您好，您的账户余额已低于一万二千七百六十六令吉五十仙，请及时充值。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "快递员将在上午9点10分左右到达，请在24-34之间留意接收。"}, {"role": "assistant", "content": "快递员将在上午九点十分左右到达，请在二十四到三十四之间留意接收。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您好，您的账户余额已低于RM127.50，请及时充值。"}, {"role": "assistant", "content": "您好，您的账户余额已低于一百二十七令吉五十仙，请及时充值。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "由于人数较多，预计等待时间为4.4小时。"}, {"role": "assistant", "content": "由于人数较多，预计等待时间为四点四小时。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您排在第8位，还需等待约68分钟。"}, {"role": "assistant", "content": "您排在第八位，还需等待约六十八分钟。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您排在第23位，还需等待约5分钟。"}, {"role": "assistant", "content": "您排在第二十三位，还需等待约五分钟。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请在19-21个工作日内等待审核结果。"}, {"role": "assistant", "content": "请在十九到二十一个工作日内等待审核结果。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "快递员将在上午8点15分左右到达，请在18–21之间留意接收。"}, {"role": "assistant", "content": "快递员将在上午八点十五分左右到达，请在十八到二十一之间留意接收。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您排在第5位，还需等待约8分钟。"}, {"role": "assistant", "content": "您排在第五位，还需等待约八分钟。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您好，您的账户余额已低于RM 277，请及时充值。"}, {"role": "assistant", "content": "您好，您的账户余额已低于二百七十七令吉，请及时充值。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请在21 - 27个工作日内等待审核结果。"}, {"role": "assistant", "content": "请在二十一到二十七个工作日内等待审核结果。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请在9-13个工作日内等待审核结果。"}, {"role": "assistant", "content": "请在九到十三个工作日内等待审核结果。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "快递员将在4点左右到达，请在1-11之间留意接收。"}, {"role": "assistant", "content": "快递员将在四点左右到达，请在一到十一之间留意接收。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您好，您的账户余额已低于RM 728，请及时充值。"}, {"role": "assistant", "content": "您好，您的账户余额已低于七百二十八令吉，请及时充值。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "由于人数较多，预计等待时间为37.37小时。"}, {"role": "assistant", "content": "由于人数较多，预计等待时间为三十七点三七小时。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "快递员将在22:20左右到达，请在4-13之间留意接收。"}, {"role": "assistant", "content": "快递员将在二十二点二十分左右到达，请在四到十三之间留意接收。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请在18-28个工作日内等待审核结果。"}, {"role": "assistant", "content": "请在十八到二十八个工作日内等待审核结果。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请在19–22个工作日内等待审核结果。"}, {"role": "assistant", "content": "请在十九到二十二个工作日内等待审核结果。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请在14 - 16个工作日内等待审核结果。"}, {"role": "assistant", "content": "请在十四到十六个工作日内等待审核结果。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您排在第5位，还需等待约15分钟。"}, {"role": "assistant", "content": "您排在第五位，还需等待约十五分钟。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "快递员将在21:00左右到达，请在11-21之间留意接收。"}, {"role": "assistant", "content": "快递员将在二十一点左右到达，请在十一到二十一之间留意接收。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "快递员将在12:00左右到达，请在8 - 17之间留意接收。"}, {"role": "assistant", "content": "快递员将在十二点左右到达，请在八到十七之间留意接收。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您好，您的账户余额已低于RM225.50，请及时充值。"}, {"role": "assistant", "content": "您好，您的账户余额已低于二百二十五令吉五十仙，请及时充值。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请在13-16个工作日内等待审核结果。"}, {"role": "assistant", "content": "请在十三到十六个工作日内等待审核结果。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "快递员将在3点左右到达，请在15 - 24之间留意接收。"}, {"role": "assistant", "content": "快递员将在三点左右到达，请在十五到二十四之间留意接收。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您排在第11位，还需等待约7分钟。"}, {"role": "assistant", "content": "您排在第十一位，还需等待约七分钟。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "由于人数较多，预计等待时间为95.8小时。"}, {"role": "assistant", "content": "由于人数较多，预计等待时间为九十五点八小时。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "由于人数较多，预计等待时间为59.07小时。"}, {"role": "assistant", "content": "由于人数较多，预计等待时间为五十九点零七小时。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "由于人数较多，预计等待时间为34.47小时。"}, {"role": "assistant", "content": "由于人数较多，预计等待时间为三十四点四七小时。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您排在第9位，还需等待约18分钟。"}, {"role": "assistant", "content": "您排在第九位，还需等待约十八分钟。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您排在第25位，还需等待约8分钟。"}, {"role": "assistant", "content": "您排在第二十五位，还需等待约八分钟。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请在6-8个工作日内等待审核结果。"}, {"role": "assistant", "content": "请在六到八个工作日内等待审核结果。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您排在第29位，还需等待约7分钟。"}, {"role": "assistant", "content": "您排在第二十九位，还需等待约七分钟。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您好，您的账户余额已低于RM37,881.59，请及时充值。"}, {"role": "assistant", "content": "您好，您的账户余额已低于三万七千八百八十一令吉五十九仙，请及时充值。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "由于人数较多，预计等待时间为6.2小时。"}, {"role": "assistant", "content": "由于人数较多，预计等待时间为六点二小时。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您好，您的账户余额已低于RM67,034.90，请及时充值。"}, {"role": "assistant", "content": "您好，您的账户余额已低于六万七千零三十四令吉九十仙，请及时充值。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请在16-17个工作日内等待审核结果。"}, {"role": "assistant", "content": "请在十六到十七个工作日内等待审核结果。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请在24 - 33个工作日内等待审核结果。"}, {"role": "assistant", "content": "请在二十四到三十三个工作日内等待审核结果。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您好，您的账户余额已低于RM 78,826，请及时充值。"}, {"role": "assistant", "content": "您好，您的账户余额已低于七万八千八百二十六令吉，请及时充值。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请在5-7个工作日内等待审核结果。"}, {"role": "assistant", "content": "请在五到七个工作日内等待审核结果。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "由于人数较多，预计等待时间为16.87小时。"}, {"role": "assistant", "content": "由于人数较多，预计等待时间为十六点八七小时。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "快递员将在6:15左右到达，请在27–30之间留意接收。"}, {"role": "assistant", "content": "快递员将在六点十五分左右到达，请在二十七到三十之间留意接收。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "由于人数较多，预计等待时间为74.03小时。"}, {"role": "assistant", "content": "由于人数较多，预计等待时间为七十四点零三小时。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您排在第9位，还需等待约41分钟。"}, {"role": "assistant", "content": "您排在第九位，还需等待约四十一分钟。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请在29 - 36个工作日内等待审核结果。"}, {"role": "assistant", "content": "请在二十九到三十六个工作日内等待审核结果。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "快递员将在下午8:40左右到达，请在26-34之间留意接收。"}, {"role": "assistant", "content": "快递员将在下午八点四十分左右到达，请在二十六到三十四之间留意接收。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请在11-21个工作日内等待审核结果。"}, {"role": "assistant", "content": "请在十一到二十一个工作日内等待审核结果。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您排在第12位，还需等待约9分钟。"}, {"role": "assistant", "content": "您排在第十二位，还需等待约九分钟。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请在24-32个工作日内等待审核结果。"}, {"role": "assistant", "content": "请在二十四到三十二个工作日内等待审核结果。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请在24-34个工作日内等待审核结果。"}, {"role": "assistant", "content": "请在二十四到三十四个工作日内等待审核结果。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请在27 - 30个工作日内等待审核结果。"}, {"role": "assistant", "content": "请在二十七到三十个工作日内等待审核结果。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "快递员将在3:20左右到达，请在16–23之间留意接收。"}, {"role": "assistant", "content": "快递员将在三点二十分左右到达，请在十六到二十三之间留意接收。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您排在第8位，还需等待约8分钟。"}, {"role": "assistant", "content": "您排在第八位，还需等待约八分钟。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请在23-25个工作日内等待审核结果。"}, {"role": "assistant", "content": "请在二十三到二十五个工作日内等待审核结果。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "由于人数较多，预计等待时间为26.20小时。"}, {"role": "assistant", "content": "由于人数较多，预计等待时间为二十六点二零小时。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您排在第7位，还需等待约9分钟。"}, {"role": "assistant", "content": "您排在第七位，还需等待约九分钟。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "由于人数较多，预计等待时间为51.27小时。"}, {"role": "assistant", "content": "由于人数较多，预计等待时间为五十一点二七小时。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您排在第17位，还需等待约888分钟。"}, {"role": "assistant", "content": "您排在第十七位，还需等待约八百八十八分钟。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您好，您的账户余额已低于RM97，请及时充值。"}, {"role": "assistant", "content": "您好，您的账户余额已低于九十七令吉，请及时充值。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "由于人数较多，预计等待时间为8.1小时。"}, {"role": "assistant", "content": "由于人数较多，预计等待时间为八点一小时。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "由于人数较多，预计等待时间为14.54小时。"}, {"role": "assistant", "content": "由于人数较多，预计等待时间为十四点五四小时。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您排在第22位，还需等待约1分钟。"}, {"role": "assistant", "content": "您排在第二十二位，还需等待约一分钟。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您排在第30位，还需等待约273分钟。"}, {"role": "assistant", "content": "您排在第三十位，还需等待约二百七十三分钟。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请在16 - 18个工作日内等待审核结果。"}, {"role": "assistant", "content": "请在十六到十八个工作日内等待审核结果。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "由于人数较多，预计等待时间为82.63小时。"}, {"role": "assistant", "content": "由于人数较多，预计等待时间为八十二点六三小时。"}], "lang": "zh", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கின் 23ஆவது மாதக் கட்டணத்தைச் செலுத்தவும்."}, {"role": "assistant", "content": "உங்கள் கணக்கின் இருபத்து மூன்றாவது மாதக் கட்டணத்தைச் செலுத்தவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் பணப் பரிமாற்றம் 11:00 am மணிக்கு வெற்றிகரமாக முடிந்தது."}, {"role": "assistant", "content": "உங்கள் பணப் பரிமாற்றம் பதினொன்று ஏ எம் மணிக்கு வெற்றிகரமாக முடிந்தது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் 013-446 1764 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பூஜ்ஜியம் ஒன்று மூன்று நான்கு நான்கு ஆறு ஒன்று ஏழு ஆறு நான்கு என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் பணப் பரிமாற்றம் 5:05 pm மணிக்கு வெற்றிகரமாக முடிந்தது."}, {"role": "assistant", "content": "உங்கள் பணப் பரிமாற்றம் ஐந்து பூஜ்ஜியம் ஐந்து பி எம் மணிக்கு வெற்றிகரமாக முடிந்தது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் 03-8344 2481 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பூஜ்ஜியம் மூன்று எட்டு மூன்று நான்கு நான்கு இரண்டு நான்கு எட்டு ஒன்று என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் பணப் பரிமாற்றம் 9:15 a.m. மணிக்கு வெற்றிகரமாக முடிந்தது."}, {"role": "assistant", "content": "உங்கள் பணப் பரிமாற்றம் ஒன்பது பதினைந்து ஏ எம் மணிக்கு வெற்றிகரமாக முடிந்தது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் பணப் பரிமாற்றம் 3a.m. மணிக்கு வெற்றிகரமாக முடிந்தது."}, {"role": "assistant", "content": "உங்கள் பணப் பரிமாற்றம் மூன்று ஏ எம் மணிக்கு வெற்றிகரமாக முடிந்தது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்தச் சேவை 2004 ஆம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}, {"role": "assistant", "content": "இந்தச் சேவை இரண்டாயிரத்து நான்காம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கில் RM605.90 வரவு வைக்கப்பட்டுள்ளது."}, {"role": "assistant", "content": "உங்கள் கணக்கில் அறுநூற்று ஐந்து ரிங்கிட் தொண்ணூறு சென் வரவு வைக்கப்பட்டுள்ளது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கின் 12வது மாதக் கட்டணத்தைச் செலுத்தவும்."}, {"role": "assistant", "content": "உங்கள் கணக்கின் பன்னிரண்டாவது மாதக் கட்டணத்தைச் செலுத்தவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்தச் சேவை 2008 ஆம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}, {"role": "assistant", "content": "இந்தச் சேவை இரண்டாயிரத்து எட்டாம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கின் 7வது மாதக் கட்டணத்தைச் செலுத்தவும்."}, {"role": "assistant", "content": "உங்கள் கணக்கின் ஏழாவது மாதக் கட்டணத்தைச் செலுத்தவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்தச் சேவை 1954 ஆம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}, {"role": "assistant", "content": "இந்தச் சேவை ஆயிரத்து தொள்ளாயிரத்து ஐம்பத்து நான்காம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் 017-59380228 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பூஜ்ஜியம் ஒன்று ஏழு ஐந்து ஒன்பது மூன்று எட்டு பூஜ்ஜியம் இரண்டு இரண்டு எட்டு என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கில் RM 95,777.90 வரவு வைக்கப்பட்டுள்ளது."}, {"role": "assistant", "content": "உங்கள் கணக்கில் தொண்ணூற்று ஐந்தாயிரத்து எழுநூற்று எழுபத்தேழு ரிங்கிட் தொண்ணூறு சென் வரவு வைக்கப்பட்டுள்ளது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் 018-29807023 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பூஜ்ஜியம் ஒன்று எட்டு இரண்டு ஒன்பது எட்டு பூஜ்ஜியம் ஏழு பூஜ்ஜியம் இரண்டு மூன்று என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் பணப் பரிமாற்றம் 11:45 மணிக்கு வெற்றிகரமாக முடிந்தது."}, {"role": "assistant", "content": "உங்கள் பணப் பரிமாற்றம் பதினொன்று நாற்பத்தைந்து மணிக்கு வெற்றிகரமாக முடிந்தது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்தச் சேவை 1990 ஆம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}, {"role": "assistant", "content": "இந்தச் சேவை ஆயிரத்து தொள்ளாயிரத்து தொண்ணூறாம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கில் RM 56.53 வரவு வைக்கப்பட்டுள்ளது."}, {"role": "assistant", "content": "உங்கள் கணக்கில் ஐம்பத்தாறு ரிங்கிட் ஐம்பத்து மூன்று சென் வரவு வைக்கப்பட்டுள்ளது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கில் RM36,238 வரவு வைக்கப்பட்டுள்ளது."}, {"role": "assistant", "content": "உங்கள் கணக்கில் முப்பத்தாறாயிரத்து இருநூற்று முப்பத்தெட்டு ரிங்கிட் வரவு வைக்கப்பட்டுள்ளது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் 017-85188419 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பூஜ்ஜியம் ஒன்று ஏழு எட்டு ஐந்து ஒன்று எட்டு எட்டு நான்கு ஒன்று ஒன்பது என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கின் 13வது மாதக் கட்டணத்தைச் செலுத்தவும்."}, {"role": "assistant", "content": "உங்கள் கணக்கின் பதிமூன்றாவது மாதக் கட்டணத்தைச் செலுத்தவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கின் 9வது மாதக் கட்டணத்தைச் செலுத்தவும்."}, {"role": "assistant", "content": "உங்கள் கணக்கின் ஒன்பதாவது மாதக் கட்டணத்தைச் செலுத்தவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் பணப் பரிமாற்றம் 7a.m. மணிக்கு வெற்றிகரமாக முடிந்தது."}, {"role": "assistant", "content": "உங்கள் பணப் பரிமாற்றம் ஏழு ஏ எம் மணிக்கு வெற்றிகரமாக முடிந்தது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கில் RM458.90 வரவு வைக்கப்பட்டுள்ளது."}, {"role": "assistant", "content": "உங்கள் கணக்கில் நானூற்று ஐம்பத்தெட்டு ரிங்கிட் தொண்ணூறு சென் வரவு வைக்கப்பட்டுள்ளது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் பணப் பரிமாற்றம் 1p.m. மணிக்கு வெற்றிகரமாக முடிந்தது."}, {"role": "assistant", "content": "உங்கள் பணப் பரிமாற்றம் ஒன்று பி எம் மணிக்கு வெற்றிகரமாக முடிந்தது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கில் RM641.90 வரவு வைக்கப்பட்டுள்ளது."}, {"role": "assistant", "content": "உங்கள் கணக்கில் அறுநூற்று நாற்பத்தொன்று ரிங்கிட் தொண்ணூறு சென் வரவு வைக்கப்பட்டுள்ளது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்தச் சேவை 1987 ஆம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}, {"role": "assistant", "content": "இந்தச் சேவை ஆயிரத்து தொள்ளாயிரத்து எண்பத்தேழாம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்தச் சேவை 1982 ஆம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}, {"role": "assistant", "content": "இந்தச் சேவை ஆயிரத்து தொள்ளாயிரத்து எண்பத்திரண்டாம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் 03-0608 7418 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பூஜ்ஜியம் மூன்று பூஜ்ஜியம் ஆறு பூஜ்ஜியம் எட்டு ஏழு நான்கு ஒன்று எட்டு என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் 03-4326 6589 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பூஜ்ஜியம் மூன்று நான்கு மூன்று இரண்டு ஆறு ஆறு ஐந்து எட்டு ஒன்பது என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் 014-70308750 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பூஜ்ஜியம் ஒன்று நான்கு ஏழு பூஜ்ஜியம் மூன்று பூஜ்ஜியம் எட்டு ஏழு ஐந்து பூஜ்ஜியம் என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கில் RM89 வரவு வைக்கப்பட்டுள்ளது."}, {"role": "assistant", "content": "உங்கள் கணக்கில் எண்பத்தொன்பது ரிங்கிட் வரவு வைக்கப்பட்டுள்ளது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் +60 17-672 4381 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பிளஸ் ஆறு பூஜ்ஜியம் ஒன்று ஏழு ஆறு ஏழு இரண்டு நான்கு மூன்று எட்டு ஒன்று என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் 03-8945 8345 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பூஜ்ஜியம் மூன்று எட்டு ஒன்பது நான்கு ஐந்து எட்டு மூன்று நான்கு ஐந்து என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் 014-26699617 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பூஜ்ஜியம் ஒன்று நான்கு இரண்டு ஆறு ஆறு ஒன்பது ஒன்பது ஆறு ஒன்று ஏழு என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கின் 22வது மாதக் கட்டணத்தைச் செலுத்தவும்."}, {"role": "assistant", "content": "உங்கள் கணக்கின் இருபத்திரண்டாவது மாதக் கட்டணத்தைச் செலுத்தவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்தச் சேவை 1997 ஆம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}, {"role": "assistant", "content": "இந்தச் சேவை ஆயிரத்து தொள்ளாயிரத்து தொண்ணூற்று ஏழாம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கின் 21ஆவது மாதக் கட்டணத்தைச் செலுத்தவும்."}, {"role": "assistant", "content": "உங்கள் கணக்கின் இருபத்தொன்றாவது மாதக் கட்டணத்தைச் செலுத்தவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கில் RM7,378.90 வரவு வைக்கப்பட்டுள்ளது."}, {"role": "assistant", "content": "உங்கள் கணக்கில் ஏழாயிரத்து முந்நூற்று எழுபத்தெட்டு ரிங்கிட் தொண்ணூறு சென் வரவு வைக்கப்பட்டுள்ளது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்தச் சேவை 2034 ஆம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}, {"role": "assistant", "content": "இந்தச் சேவை இரண்டாயிரத்து முப்பத்து நான்காம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் 03-4918 4417 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பூஜ்ஜியம் மூன்று நான்கு ஒன்பது ஒன்று எட்டு நான்கு நான்கு ஒன்று ஏழு என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் 018-248 9928 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பூஜ்ஜியம் ஒன்று எட்டு இரண்டு நான்கு எட்டு ஒன்பது ஒன்பது இரண்டு எட்டு என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கின் 11வது மாதக் கட்டணத்தைச் செலுத்தவும்."}, {"role": "assistant", "content": "உங்கள் கணக்கின் பதினொன்றாவது மாதக் கட்டணத்தைச் செலுத்தவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் 03-8762 9532 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பூஜ்ஜியம் மூன்று எட்டு ஏழு ஆறு இரண்டு ஒன்பது ஐந்து மூன்று இரண்டு என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கில் RM 15,550 வரவு வைக்கப்பட்டுள்ளது."}, {"role": "assistant", "content": "உங்கள் கணக்கில் பதினைந்தாயிரத்து ஐந்நூற்று ஐம்பது ரிங்கிட் வரவு வைக்கப்பட்டுள்ளது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்தச் சேவை 1998 ஆம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}, {"role": "assistant", "content": "இந்தச் சேவை ஆயிரத்து தொள்ளாயிரத்து தொண்ணூற்று எட்டாம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்தச் சேவை 2002 ஆம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}, {"role": "assistant", "content": "இந்தச் சேவை இரண்டாயிரத்து இரண்டாம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கில் RM322 வரவு வைக்கப்பட்டுள்ளது."}, {"role": "assistant", "content": "உங்கள் கணக்கில் முந்நூற்று இருபத்திரண்டு ரிங்கிட் வரவு வைக்கப்பட்டுள்ளது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் 012-940 3837 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பூஜ்ஜியம் ஒன்று இரண்டு ஒன்பது நான்கு பூஜ்ஜியம் மூன்று எட்டு மூன்று ஏழு என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கில் RM 359.39 வரவு வைக்கப்பட்டுள்ளது."}, {"role": "assistant", "content": "உங்கள் கணக்கில் முந்நூற்று ஐம்பத்தொன்பது ரிங்கிட் முப்பத்தொன்பது சென் வரவு வைக்கப்பட்டுள்ளது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் +60 19-169 4890 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பிளஸ் ஆறு பூஜ்ஜியம் ஒன்று ஒன்பது ஒன்று ஆறு ஒன்பது நான்கு எட்டு ஒன்பது பூஜ்ஜியம் என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் 03-7953 3825 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பூஜ்ஜியம் மூன்று ஏழு ஒன்பது ஐந்து மூன்று மூன்று எட்டு இரண்டு ஐந்து என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் பணப் பரிமாற்றம் 10:20 AM மணிக்கு வெற்றிகரமாக முடிந்தது."}, {"role": "assistant", "content": "உங்கள் பணப் பரிமாற்றம் பத்து இருபது ஏ எம் மணிக்கு வெற்றிகரமாக முடிந்தது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் பணப் பரிமாற்றம் 2am மணிக்கு வெற்றிகரமாக முடிந்தது."}, {"role": "assistant", "content": "உங்கள் பணப் பரிமாற்றம் இரண்டு ஏ எம் மணிக்கு வெற்றிகரமாக முடிந்தது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கில் RM 73,136 வரவு வைக்கப்பட்டுள்ளது."}, {"role": "assistant", "content": "உங்கள் கணக்கில் எழுபத்து மூன்றாயிரத்து நூற்று முப்பத்தாறு ரிங்கிட் வரவு வைக்கப்பட்டுள்ளது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கின் 23வது மாதக் கட்டணத்தைச் செலுத்தவும்."}, {"role": "assistant", "content": "உங்கள் கணக்கின் இருபத்து மூன்றாவது மாதக் கட்டணத்தைச் செலுத்தவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் +60 19-189 8181 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பிளஸ் ஆறு பூஜ்ஜியம் ஒன்று ஒன்பது ஒன்று எட்டு ஒன்பது எட்டு ஒன்று எட்டு ஒன்று என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்தச் சேவை 2023 ஆம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}, {"role": "assistant", "content": "இந்தச் சேவை இரண்டாயிரத்து இருபத்து மூன்றாம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கின் 10வது மாதக் கட்டணத்தைச் செலுத்தவும்."}, {"role": "assistant", "content": "உங்கள் கணக்கின் பத்தாவது மாதக் கட்டணத்தைச் செலுத்தவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கின் 26வது மாதக் கட்டணத்தைச் செலுத்தவும்."}, {"role": "assistant", "content": "உங்கள் கணக்கின் இருபத்தாறாவது மாதக் கட்டணத்தைச் செலுத்தவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் பணப் பரிமாற்றம் 7:40 மணிக்கு வெற்றிகரமாக முடிந்தது."}, {"role": "assistant", "content": "உங்கள் பணப் பரிமாற்றம் ஏழு நாற்பது மணிக்கு வெற்றிகரமாக முடிந்தது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கின் 6வது மாதக் கட்டணத்தைச் செலுத்தவும்."}, {"role": "assistant", "content": "உங்கள் கணக்கின் ஆறாவது மாதக் கட்டணத்தைச் செலுத்தவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்தச் சேவை 1951 ஆம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}, {"role": "assistant", "content": "இந்தச் சேவை ஆயிரத்து தொள்ளாயிரத்து ஐம்பத்தொன்றாம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் பணப் பரிமாற்றம் 8AM மணிக்கு வெற்றிகரமாக முடிந்தது."}, {"role": "assistant", "content": "உங்கள் பணப் பரிமாற்றம் எட்டு ஏ எம் மணிக்கு வெற்றிகரமாக முடிந்தது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்தச் சேவை 1952 ஆம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}, {"role": "assistant", "content": "இந்தச் சேவை ஆயிரத்து தொள்ளாயிரத்து ஐம்பத்திரண்டாம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் +60 19-110 6523 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பிளஸ் ஆறு பூஜ்ஜியம் ஒன்று ஒன்பது ஒன்று ஒன்று பூஜ்ஜியம் ஆறு ஐந்து இரண்டு மூன்று என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கில் RM93,674.85 வரவு வைக்கப்பட்டுள்ளது."}, {"role": "assistant", "content": "உங்கள் கணக்கில் தொண்ணூற்று மூன்றாயிரத்து அறுநூற்று எழுபத்து நான்கு ரிங்கிட் எண்பத்தைந்து சென் வரவு வைக்கப்பட்டுள்ளது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கில் RM 712.90 வரவு வைக்கப்பட்டுள்ளது."}, {"role": "assistant", "content": "உங்கள் கணக்கில் எழுநூற்று பன்னிரண்டு ரிங்கிட் தொண்ணூறு சென் வரவு வைக்கப்பட்டுள்ளது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கில் RM360 வரவு வைக்கப்பட்டுள்ளது."}, {"role": "assistant", "content": "உங்கள் கணக்கில் முந்நூற்று அறுபது ரிங்கிட் வரவு வைக்கப்பட்டுள்ளது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கில் RM 10,813.90 வரவு வைக்கப்பட்டுள்ளது."}, {"role": "assistant", "content": "உங்கள் கணக்கில் பத்தாயிரத்து எண்ணூற்று பதிமூன்று ரிங்கிட் தொண்ணூறு சென் வரவு வைக்கப்பட்டுள்ளது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் பணப் பரிமாற்றம் 11:45 PM மணிக்கு வெற்றிகரமாக முடிந்தது."}, {"role": "assistant", "content": "உங்கள் பணப் பரிமாற்றம் பதினொன்று நாற்பத்தைந்து பி எம் மணிக்கு வெற்றிகரமாக முடிந்தது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் 010-932 2628 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பூஜ்ஜியம் ஒன்று பூஜ்ஜியம் ஒன்பது மூன்று இரண்டு இரண்டு ஆறு இரண்டு எட்டு என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் 03-5331 5466 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பூஜ்ஜியம் மூன்று ஐந்து மூன்று மூன்று ஒன்று ஐந்து நான்கு ஆறு ஆறு என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் பணப் பரிமாற்றம் 15:30 மணிக்கு வெற்றிகரமாக முடிந்தது."}, {"role": "assistant", "content": "உங்கள் பணப் பரிமாற்றம் பதினைந்து முப்பது மணிக்கு வெற்றிகரமாக முடிந்தது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் பணப் பரிமாற்றம் 5.00 மணிக்கு வெற்றிகரமாக முடிந்தது."}, {"role": "assistant", "content": "உங்கள் பணப் பரிமாற்றம் ஐந்து மணிக்கு வெற்றிகரமாக முடிந்தது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் பணப் பரிமாற்றம் 9.40 மணிக்கு வெற்றிகரமாக முடிந்தது."}, {"role": "assistant", "content": "உங்கள் பணப் பரிமாற்றம் ஒன்பது நாற்பது மணிக்கு வெற்றிகரமாக முடிந்தது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் பணப் பரிமாற்றம் 7PM மணிக்கு வெற்றிகரமாக முடிந்தது."}, {"role": "assistant", "content": "உங்கள் பணப் பரிமாற்றம் ஏழு பி எம் மணிக்கு வெற்றிகரமாக முடிந்தது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் 015-38210274 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பூஜ்ஜியம் ஒன்று ஐந்து மூன்று எட்டு இரண்டு ஒன்று பூஜ்ஜியம் இரண்டு ஏழு நான்கு என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் பணப் பரிமாற்றம் 3.00 மணிக்கு வெற்றிகரமாக முடிந்தது."}, {"role": "assistant", "content": "உங்கள் பணப் பரிமாற்றம் மூன்று மணிக்கு வெற்றிகரமாக முடிந்தது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கின் 1ஆவது மாதக் கட்டணத்தைச் செலுத்தவும்."}, {"role": "assistant", "content": "உங்கள் கணக்கின் முதலாவது மாதக் கட்டணத்தைச் செலுத்தவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கின் 24ஆவது மாதக் கட்டணத்தைச் செலுத்தவும்."}, {"role": "assistant", "content": "உங்கள் கணக்கின் இருபத்து நான்காவது மாதக் கட்டணத்தைச் செலுத்தவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கின் 26ஆவது மாதக் கட்டணத்தைச் செலுத்தவும்."}, {"role": "assistant", "content": "உங்கள் கணக்கின் இருபத்தாறாவது மாதக் கட்டணத்தைச் செலுத்தவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் +60 11-674 1273 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பிளஸ் ஆறு பூஜ்ஜியம் ஒன்று ஒன்று ஆறு ஏழு நான்கு ஒன்று இரண்டு ஏழு மூன்று என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கின் 5ஆவது மாதக் கட்டணத்தைச் செலுத்தவும்."}, {"role": "assistant", "content": "உங்கள் கணக்கின் ஐந்தாவது மாதக் கட்டணத்தைச் செலுத்தவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கில் RM849.90 வரவு வைக்கப்பட்டுள்ளது."}, {"role": "assistant", "content": "உங்கள் கணக்கில் எண்ணூற்று நாற்பத்தொன்பது ரிங்கிட் தொண்ணூறு சென் வரவு வைக்கப்பட்டுள்ளது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கில் RM960.50 வரவு வைக்கப்பட்டுள்ளது."}, {"role": "assistant", "content": "உங்கள் கணக்கில் தொள்ளாயிரத்து அறுபது ரிங்கிட் ஐம்பது சென் வரவு வைக்கப்பட்டுள்ளது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்தச் சேவை 2026 ஆம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}, {"role": "assistant", "content": "இந்தச் சேவை இரண்டாயிரத்து இருபத்தாறாம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் +60 15-214 8353 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பிளஸ் ஆறு பூஜ்ஜியம் ஒன்று ஐந்து இரண்டு ஒன்று நான்கு எட்டு மூன்று ஐந்து மூன்று என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் பணப் பரிமாற்றம் 15:05 மணிக்கு வெற்றிகரமாக முடிந்தது."}, {"role": "assistant", "content": "உங்கள் பணப் பரிமாற்றம் பதினைந்து பூஜ்ஜியம் ஐந்து மணிக்கு வெற்றிகரமாக முடிந்தது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் பணப் பரிமாற்றம் 7.10 மணிக்கு வெற்றிகரமாக முடிந்தது."}, {"role": "assistant", "content": "உங்கள் பணப் பரிமாற்றம் ஏழு பத்து மணிக்கு வெற்றிகரமாக முடிந்தது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கின் 11ஆவது மாதக் கட்டணத்தைச் செலுத்தவும்."}, {"role": "assistant", "content": "உங்கள் கணக்கின் பதினொன்றாவது மாதக் கட்டணத்தைச் செலுத்தவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் 014-395 2188 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பூஜ்ஜியம் ஒன்று நான்கு மூன்று ஒன்பது ஐந்து இரண்டு ஒன்று எட்டு எட்டு என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கில் RM590.50 வரவு வைக்கப்பட்டுள்ளது."}, {"role": "assistant", "content": "உங்கள் கணக்கில் ஐந்நூற்று தொண்ணூறு ரிங்கிட் ஐம்பது சென் வரவு வைக்கப்பட்டுள்ளது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் பணப் பரிமாற்றம் 3.30 மணிக்கு வெற்றிகரமாக முடிந்தது."}, {"role": "assistant", "content": "உங்கள் பணப் பரிமாற்றம் மூன்று முப்பது மணிக்கு வெற்றிகரமாக முடிந்தது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கில் RM 567 வரவு வைக்கப்பட்டுள்ளது."}, {"role": "assistant", "content": "உங்கள் கணக்கில் ஐந்நூற்று அறுபத்தேழு ரிங்கிட் வரவு வைக்கப்பட்டுள்ளது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்தச் சேவை 2029 ஆம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}, {"role": "assistant", "content": "இந்தச் சேவை இரண்டாயிரத்து இருபத்தொன்பதாம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கில் RM 75,677.50 வரவு வைக்கப்பட்டுள்ளது."}, {"role": "assistant", "content": "உங்கள் கணக்கில் எழுபத்தைந்தாயிரத்து அறுநூற்று எழுபத்தேழு ரிங்கிட் ஐம்பது சென் வரவு வைக்கப்பட்டுள்ளது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கின் 16வது மாதக் கட்டணத்தைச் செலுத்தவும்."}, {"role": "assistant", "content": "உங்கள் கணக்கின் பதினாறாவது மாதக் கட்டணத்தைச் செலுத்தவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கின் 21வது மாதக் கட்டணத்தைச் செலுத்தவும்."}, {"role": "assistant", "content": "உங்கள் கணக்கின் இருபத்தொன்றாவது மாதக் கட்டணத்தைச் செலுத்தவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் பணப் பரிமாற்றம் 12.30 மணிக்கு வெற்றிகரமாக முடிந்தது."}, {"role": "assistant", "content": "உங்கள் பணப் பரிமாற்றம் பன்னிரண்டு முப்பது மணிக்கு வெற்றிகரமாக முடிந்தது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்தச் சேவை 1988 ஆம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}, {"role": "assistant", "content": "இந்தச் சேவை ஆயிரத்து தொள்ளாயிரத்து எண்பத்தெட்டாம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் 03-7711 8198 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பூஜ்ஜியம் மூன்று ஏழு ஏழு ஒன்று ஒன்று எட்டு ஒன்று ஒன்பது எட்டு என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் +60 17-226 0612 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பிளஸ் ஆறு பூஜ்ஜியம் ஒன்று ஏழு இரண்டு இரண்டு ஆறு பூஜ்ஜியம் ஆறு ஒன்று இரண்டு என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கில் RM44 வரவு வைக்கப்பட்டுள்ளது."}, {"role": "assistant", "content": "உங்கள் கணக்கில் நாற்பத்து நான்கு ரிங்கிட் வரவு வைக்கப்பட்டுள்ளது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் 019-78206812 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பூஜ்ஜியம் ஒன்று ஒன்பது ஏழு எட்டு இரண்டு பூஜ்ஜியம் ஆறு எட்டு ஒன்று இரண்டு என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் பணப் பரிமாற்றம் 9pm மணிக்கு வெற்றிகரமாக முடிந்தது."}, {"role": "assistant", "content": "உங்கள் பணப் பரிமாற்றம் ஒன்பது பி எம் மணிக்கு வெற்றிகரமாக முடிந்தது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் 011-69832132 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பூஜ்ஜியம் ஒன்று ஒன்று ஆறு ஒன்பது எட்டு மூன்று இரண்டு ஒன்று மூன்று இரண்டு என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கில் RM833.50 வரவு வைக்கப்பட்டுள்ளது."}, {"role": "assistant", "content": "உங்கள் கணக்கில் எண்ணூற்று முப்பத்து மூன்று ரிங்கிட் ஐம்பது சென் வரவு வைக்கப்பட்டுள்ளது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் பணப் பரிமாற்றம் 19:50 மணிக்கு வெற்றிகரமாக முடிந்தது."}, {"role": "assistant", "content": "உங்கள் பணப் பரிமாற்றம் பத்தொன்பது ஐம்பது மணிக்கு வெற்றிகரமாக முடிந்தது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கின் 19ஆவது மாதக் கட்டணத்தைச் செலுத்தவும்."}, {"role": "assistant", "content": "உங்கள் கணக்கின் பத்தொன்பதாவது மாதக் கட்டணத்தைச் செலுத்தவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் பணப் பரிமாற்றம் 11.05 மணிக்கு வெற்றிகரமாக முடிந்தது."}, {"role": "assistant", "content": "உங்கள் பணப் பரிமாற்றம் பதினொன்று பூஜ்ஜியம் ஐந்து மணிக்கு வெற்றிகரமாக முடிந்தது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கின் 12ஆவது மாதக் கட்டணத்தைச் செலுத்தவும்."}, {"role": "assistant", "content": "உங்கள் கணக்கின் பன்னிரண்டாவது மாதக் கட்டணத்தைச் செலுத்தவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கில் RM 67.90 வரவு வைக்கப்பட்டுள்ளது."}, {"role": "assistant", "content": "உங்கள் கணக்கில் அறுபத்தேழு ரிங்கிட் தொண்ணூறு சென் வரவு வைக்கப்பட்டுள்ளது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கின் 22ஆவது மாதக் கட்டணத்தைச் செலுத்தவும்."}, {"role": "assistant", "content": "உங்கள் கணக்கின் இருபத்திரண்டாவது மாதக் கட்டணத்தைச் செலுத்தவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் 018-365 6881 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பூஜ்ஜியம் ஒன்று எட்டு மூன்று ஆறு ஐந்து ஆறு எட்டு எட்டு ஒன்று என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கில் RM 4 வரவு வைக்கப்பட்டுள்ளது."}, {"role": "assistant", "content": "உங்கள் கணக்கில் நான்கு ரிங்கிட் வரவு வைக்கப்பட்டுள்ளது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் பணப் பரிமாற்றம் 7:30 மணிக்கு வெற்றிகரமாக முடிந்தது."}, {"role": "assistant", "content": "உங்கள் பணப் பரிமாற்றம் ஏழு முப்பது மணிக்கு வெற்றிகரமாக முடிந்தது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்தச் சேவை 1976 ஆம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}, {"role": "assistant", "content": "இந்தச் சேவை ஆயிரத்து தொள்ளாயிரத்து எழுபத்தாறாம் ஆண்டு முதல் அறிமுகப்படுத்தப்பட்டது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் 013-55005879 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பூஜ்ஜியம் ஒன்று மூன்று ஐந்து ஐந்து பூஜ்ஜியம் பூஜ்ஜியம் ஐந்து எட்டு ஏழு ஒன்பது என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் பணப் பரிமாற்றம் 7:30 pm மணிக்கு வெற்றிகரமாக முடிந்தது."}, {"role": "assistant", "content": "உங்கள் பணப் பரிமாற்றம் ஏழு முப்பது பி எம் மணிக்கு வெற்றிகரமாக முடிந்தது."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் +60 13-173 6077 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பிளஸ் ஆறு பூஜ்ஜியம் ஒன்று மூன்று ஒன்று ஏழு மூன்று ஆறு பூஜ்ஜியம் ஏழு ஏழு என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் 03-8495 8365 என்ற எண்ணிற்கு அழைக்கவும்."}, {"role": "assistant", "content": "உங்களுக்கு ஏதேனும் சந்தேகம் இருந்தால் பூஜ்ஜியம் மூன்று எட்டு நான்கு ஒன்பது ஐந்து எட்டு மூன்று ஆறு ஐந்து என்ற எண்ணிற்கு அழைக்கவும்."}], "lang": "ta", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை LKR 41.99."}, {"role": "assistant", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை நாற்பத்தொன்று ரூபாய் தொண்ணூற்று ஒன்பது சதம்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 25 kg இணைய வசதியை ரூ. 83 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் இருபத்தைந்து கிலோகிராம் இணைய வசதியை எண்பத்து மூன்று ரூபாய் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை LKR 79,819.50."}, {"role": "assistant", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை எழுபத்தொன்பதாயிரத்து எண்ணூற்று பத்தொன்பது ரூபாய் ஐம்பது சதம்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை ரூ. 781.90."}, {"role": "assistant", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை எழுநூற்று எண்பத்தொன்று ரூபாய் தொண்ணூறு சதம்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 80.2ml இணைய வசதியை Rs.825.57 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் எண்பது புள்ளி இரண்டு மில்லிலிட்டர் இணைய வசதியை எண்ணூற்று இருபத்தைந்து ரூபாய் ஐம்பத்தேழு சதம் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை 2ஆவது மாடியில் நடைபெறும்."}, {"role": "assistant", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை இரண்டாவது மாடியில் நடைபெறும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 69.9 l இணைய வசதியை Rs. 321 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் அறுபத்தொன்பது புள்ளி ஒன்பது லிட்டர் இணைய வசதியை முந்நூற்று இருபத்தொன்று ரூபாய் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உடல் வெப்ப நிலை 57.3GB ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}, {"role": "assistant", "content": "உடல் வெப்ப நிலை ஐம்பத்தேழு புள்ளி மூன்று ஜிபி ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "மேலதிக விபரங்களுக்கு https://example.com/faq என்ற இணையதளத்தைப் பார்க்கவும்."}, {"role": "assistant", "content": "மேலதிக விபரங்களுக்கு h t t p s colon slash slash example dot com slash faq என்ற இணையதளத்தைப் பார்க்கவும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 488°C இணைய வசதியை ரூ. 86 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் நானூற்று எண்பத்தெட்டு டிகிரி செல்சியஸ் இணைய வசதியை எண்பத்தாறு ரூபாய் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "மேலதிக விபரங்களுக்கு https://gmail.co.id என்ற இணையதளத்தைப் பார்க்கவும்."}, {"role": "assistant", "content": "மேலதிக விபரங்களுக்கு h t t p s colon slash slash gmail dot co dot id என்ற இணையதளத்தைப் பார்க்கவும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 314kg இணைய வசதியை ரூ. 22,101.50 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் முந்நூற்று பதினான்கு கிலோகிராம் இணைய வசதியை இருபத்திரண்டாயிரத்து நூற்று ஒன்று ரூபாய் ஐம்பது சதம் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை LKR 31,221.97."}, {"role": "assistant", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை முப்பத்தொன்றாயிரத்து இருநூற்று இருபத்தொன்று ரூபாய் தொண்ணூற்று ஏழு சதம்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உடல் வெப்ப நிலை 47GB ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}, {"role": "assistant", "content": "உடல் வெப்ப நிலை நாற்பத்தேழு ஜிபி ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 495l இணைய வசதியை LKR 1,268.79 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் நானூற்று தொண்ணூற்று ஐந்து லிட்டர் இணைய வசதியை ஆயிரத்து இருநூற்று அறுபத்தெட்டு ரூபாய் எழுபத்தொன்பது சதம் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை Rs.265.90."}, {"role": "assistant", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை இருநூற்று அறுபத்தைந்து ரூபாய் தொண்ணூறு சதம்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை 19ஆவது மாடியில் நடைபெறும்."}, {"role": "assistant", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை பத்தொன்பதாவது மாடியில் நடைபெறும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை 1வது மாடியில் நடைபெறும்."}, {"role": "assistant", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை முதலாவது மாடியில் நடைபெறும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை Rs. 64,969.50."}, {"role": "assistant", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை அறுபத்து நான்காயிரத்து தொள்ளாயிரத்து அறுபத்தொன்பது ரூபாய் ஐம்பது சதம்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உடல் வெப்ப நிலை 436GB ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}, {"role": "assistant", "content": "உடல் வெப்ப நிலை நானூற்று முப்பத்தாறு ஜிபி ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 271kg இணைய வசதியை ரூ. 38,001 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் இருநூற்று எழுபத்தொன்று கிலோகிராம் இணைய வசதியை முப்பத்தெட்டாயிரத்து ஒன்று ரூபாய் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை Rs.62,831.50."}, {"role": "assistant", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை அறுபத்திரண்டாயிரத்து எண்ணூற்று முப்பத்தொன்று ரூபாய் ஐம்பது சதம்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை 7ஆவது மாடியில் நடைபெறும்."}, {"role": "assistant", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை ஏழாவது மாடியில் நடைபெறும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை Rs.45.50."}, {"role": "assistant", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை நாற்பத்தைந்து ரூபாய் ஐம்பது சதம்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 474°C இணைய வசதியை ரூ. 77.59 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் நானூற்று எழுபத்து நான்கு டிகிரி செல்சியஸ் இணைய வசதியை எழுபத்தேழு ரூபாய் ஐம்பத்தொன்பது சதம் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 1km இணைய வசதியை ரூ. 78.90 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் ஒன்று கிலோமீட்டர் இணைய வசதியை எழுபத்தெட்டு ரூபாய் தொண்ணூறு சதம் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 14.2 °C இணைய வசதியை LKR 57.29 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் பதினான்கு புள்ளி இரண்டு டிகிரி செல்சியஸ் இணைய வசதியை ஐம்பத்தேழு ரூபாய் இருபத்தொன்பது சதம் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 46.3 cm இணைய வசதியை Rs.23,335.64 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் நாற்பத்தாறு புள்ளி மூன்று சென்டிமீட்டர் இணைய வசதியை இருபத்து மூன்றாயிரத்து முந்நூற்று முப்பத்தைந்து ரூபாய் அறுபத்து நான்கு சதம் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 27 GB இணைய வசதியை LKR 18,951.50 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் இருபத்தேழு ஜிபி இணைய வசதியை பதினெட்டாயிரத்து தொள்ளாயிரத்து ஐம்பத்தொன்று ரூபாய் ஐம்பது சதம் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "மேலதிக விபரங்களுக்கு https://posta.it/help என்ற இணையதளத்தைப் பார்க்கவும்."}, {"role": "assistant", "content": "மேலதிக விபரங்களுக்கு h t t p s colon slash slash posta dot it slash help என்ற இணையதளத்தைப் பார்க்கவும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உடல் வெப்ப நிலை 81.9 ml ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}, {"role": "assistant", "content": "உடல் வெப்ப நிலை எண்பத்தொன்று புள்ளி ஒன்பது மில்லிலிட்டர் ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை 13ஆவது மாடியில் நடைபெறும்."}, {"role": "assistant", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை பதிமூன்றாவது மாடியில் நடைபெறும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உடல் வெப்ப நிலை 332l ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}, {"role": "assistant", "content": "உடல் வெப்ப நிலை முந்நூற்று முப்பத்திரண்டு லிட்டர் ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை LKR 424.55."}, {"role": "assistant", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை நானூற்று இருபத்து நான்கு ரூபாய் ஐம்பத்தைந்து சதம்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை ரூ. 205."}, {"role": "assistant", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை இருநூற்று ஐந்து ரூபாய்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 31km இணைய வசதியை LKR 48.50 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் முப்பத்தொன்று கிலோமீட்டர் இணைய வசதியை நாற்பத்தெட்டு ரூபாய் ஐம்பது சதம் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "மேலதிக விபரங்களுக்கு www.correo.es என்ற இணையதளத்தைப் பார்க்கவும்."}, {"role": "assistant", "content": "மேலதிக விபரங்களுக்கு w w w dot correo dot es என்ற இணையதளத்தைப் பார்க்கவும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 302 °C இணைய வசதியை ரூ. 172.90 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் முந்நூற்று இரண்டு டிகிரி செல்சியஸ் இணைய வசதியை நூற்று எழுபத்திரண்டு ரூபாய் தொண்ணூறு சதம் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 77.2GB இணைய வசதியை ரூ. 72 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் எழுபத்தேழு புள்ளி இரண்டு ஜிபி இணைய வசதியை எழுபத்திரண்டு ரூபாய் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை Rs. 59.50."}, {"role": "assistant", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை ஐம்பத்தொன்பது ரூபாய் ஐம்பது சதம்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "மேலதிக விபரங்களுக்கு www.onet.pl/login என்ற இணையதளத்தைப் பார்க்கவும்."}, {"role": "assistant", "content": "மேலதிக விபரங்களுக்கு w w w dot onet dot pl slash login என்ற இணையதளத்தைப் பார்க்கவும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உடல் வெப்ப நிலை 319km ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}, {"role": "assistant", "content": "உடல் வெப்ப நிலை முந்நூற்று பத்தொன்பது கிலோமீட்டர் ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உடல் வெப்ப நிலை 49GB ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}, {"role": "assistant", "content": "உடல் வெப்ப நிலை நாற்பத்தொன்பது ஜிபி ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை 5ஆவது மாடியில் நடைபெறும்."}, {"role": "assistant", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை ஐந்தாவது மாடியில் நடைபெறும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை Rs.48,008.90."}, {"role": "assistant", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை நாற்பத்தெட்டாயிரத்து எட்டு ரூபாய் தொண்ணூறு சதம்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உடல் வெப்ப நிலை 166 ml ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}, {"role": "assistant", "content": "உடல் வெப்ப நிலை நூற்று அறுபத்தாறு மில்லிலிட்டர் ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உடல் வெப்ப நிலை 20.9g ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}, {"role": "assistant", "content": "உடல் வெப்ப நிலை இருபது புள்ளி ஒன்பது கிராம் ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உடல் வெப்ப நிலை 1.9kg ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}, {"role": "assistant", "content": "உடல் வெப்ப நிலை ஒரு புள்ளி ஒன்பது கிலோகிராம் ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை ரூ. 117.90."}, {"role": "assistant", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை நூற்று பதினேழு ரூபாய் தொண்ணூறு சதம்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 107 GB இணைய வசதியை LKR 4,555.74 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் நூற்று ஏழு ஜிபி இணைய வசதியை நான்காயிரத்து ஐந்நூற்று ஐம்பத்தைந்து ரூபாய் எழுபத்து நான்கு சதம் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 62.6ml இணைய வசதியை ரூ. 48.41 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் அறுபத்திரண்டு புள்ளி ஆறு மில்லிலிட்டர் இணைய வசதியை நாற்பத்தெட்டு ரூபாய் நாற்பத்தொன்று சதம் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உடல் வெப்ப நிலை 35 l ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}, {"role": "assistant", "content": "உடல் வெப்ப நிலை முப்பத்தைந்து லிட்டர் ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை Rs.83.90."}, {"role": "assistant", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை எண்பத்து மூன்று ரூபாய் தொண்ணூறு சதம்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 34.3 ml இணைய வசதியை Rs. 4,680 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் முப்பத்து நான்கு புள்ளி மூன்று மில்லிலிட்டர் இணைய வசதியை நான்காயிரத்து அறுநூற்று எண்பது ரூபாய் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உடல் வெப்ப நிலை 37.4km ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}, {"role": "assistant", "content": "உடல் வெப்ப நிலை முப்பத்தேழு புள்ளி நான்கு கிலோமீட்டர் ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "மேலதிக விபரங்களுக்கு www.gmail.com என்ற இணையதளத்தைப் பார்க்கவும்."}, {"role": "assistant", "content": "மேலதிக விபரங்களுக்கு w w w dot gmail dot com என்ற இணையதளத்தைப் பார்க்கவும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "மேலதிக விபரங்களுக்கு https://scicom.com.my/help என்ற இணையதளத்தைப் பார்க்கவும்."}, {"role": "assistant", "content": "மேலதிக விபரங்களுக்கு h t t p s colon slash slash scicom dot com dot my slash help என்ற இணையதளத்தைப் பார்க்கவும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை 19வது மாடியில் நடைபெறும்."}, {"role": "assistant", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை பத்தொன்பதாவது மாடியில் நடைபெறும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை LKR 689.50."}, {"role": "assistant", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை அறுநூற்று எண்பத்தொன்பது ரூபாய் ஐம்பது சதம்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 292 GB இணைய வசதியை Rs.96 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் இருநூற்று தொண்ணூற்று இரண்டு ஜிபி இணைய வசதியை தொண்ணூற்று ஆறு ரூபாய் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 158 cm இணைய வசதியை Rs.496.90 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் நூற்று ஐம்பத்தெட்டு சென்டிமீட்டர் இணைய வசதியை நானூற்று தொண்ணூற்று ஆறு ரூபாய் தொண்ணூறு சதம் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 268km இணைய வசதியை Rs. 47 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் இருநூற்று அறுபத்தெட்டு கிலோமீட்டர் இணைய வசதியை நாற்பத்தேழு ரூபாய் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உடல் வெப்ப நிலை 9.7ml ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}, {"role": "assistant", "content": "உடல் வெப்ப நிலை ஒன்பது புள்ளி ஏழு மில்லிலிட்டர் ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை ரூ. 94,840."}, {"role": "assistant", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை தொண்ணூற்று நான்காயிரத்து எண்ணூற்று நாற்பது ரூபாய்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை ரூ. 89,251.90."}, {"role": "assistant", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை எண்பத்தொன்பதாயிரத்து இருநூற்று ஐம்பத்தொன்று ரூபாய் தொண்ணூறு சதம்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 174ml இணைய வசதியை Rs.982.50 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் நூற்று எழுபத்து நான்கு மில்லிலிட்டர் இணைய வசதியை தொள்ளாயிரத்து எண்பத்திரண்டு ரூபாய் ஐம்பது சதம் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை Rs.43,500.90."}, {"role": "assistant", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை நாற்பத்து மூன்றாயிரத்து ஐந்நூறு ரூபாய் தொண்ணூறு சதம்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை Rs. 547.50."}, {"role": "assistant", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை ஐந்நூற்று நாற்பத்தேழு ரூபாய் ஐம்பது சதம்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 367kg இணைய வசதியை LKR 99,658.48 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் முந்நூற்று அறுபத்தேழு கிலோகிராம் இணைய வசதியை தொண்ணூற்று ஒன்பதாயிரத்து அறுநூற்று ஐம்பத்தெட்டு ரூபாய் நாற்பத்தெட்டு சதம் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை 18வது மாடியில் நடைபெறும்."}, {"role": "assistant", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை பதினெட்டாவது மாடியில் நடைபெறும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உடல் வெப்ப நிலை 39.3 l ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}, {"role": "assistant", "content": "உடல் வெப்ப நிலை முப்பத்தொன்பது புள்ளி மூன்று லிட்டர் ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை 28ஆவது மாடியில் நடைபெறும்."}, {"role": "assistant", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை இருபத்தெட்டாவது மாடியில் நடைபெறும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உடல் வெப்ப நிலை 30.7 g ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}, {"role": "assistant", "content": "உடல் வெப்ப நிலை முப்பது புள்ளி ஏழு கிராம் ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "மேலதிக விபரங்களுக்கு https://www.yahoo.com.ph/support என்ற இணையதளத்தைப் பார்க்கவும்."}, {"role": "assistant", "content": "மேலதிக விபரங்களுக்கு h t t p s colon slash slash www dot yahoo dot com dot ph slash support என்ற இணையதளத்தைப் பார்க்கவும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உடல் வெப்ப நிலை 38 °C ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}, {"role": "assistant", "content": "உடல் வெப்ப நிலை முப்பத்தெட்டு டிகிரி செல்சியஸ் ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உடல் வெப்ப நிலை 352 ml ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}, {"role": "assistant", "content": "உடல் வெப்ப நிலை முந்நூற்று ஐம்பத்திரண்டு மில்லிலிட்டர் ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை 30வது மாடியில் நடைபெறும்."}, {"role": "assistant", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை முப்பதாவது மாடியில் நடைபெறும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 344 ml இணைய வசதியை Rs. 297 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் முந்நூற்று நாற்பத்து நான்கு மில்லிலிட்டர் இணைய வசதியை இருநூற்று தொண்ணூற்று ஏழு ரூபாய் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உடல் வெப்ப நிலை 99.4°C ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}, {"role": "assistant", "content": "உடல் வெப்ப நிலை தொண்ணூற்று ஒன்பது புள்ளி நான்கு டிகிரி செல்சியஸ் ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை 8ஆவது மாடியில் நடைபெறும்."}, {"role": "assistant", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை எட்டாவது மாடியில் நடைபெறும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை 4ஆவது மாடியில் நடைபெறும்."}, {"role": "assistant", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை நான்காவது மாடியில் நடைபெறும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "மேலதிக விபரங்களுக்கு www.outlook.sa/login என்ற இணையதளத்தைப் பார்க்கவும்."}, {"role": "assistant", "content": "மேலதிக விபரங்களுக்கு w w w dot outlook dot sa slash login என்ற இணையதளத்தைப் பார்க்கவும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உடல் வெப்ப நிலை 25.8cm ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}, {"role": "assistant", "content": "உடல் வெப்ப நிலை இருபத்தைந்து புள்ளி எட்டு சென்டிமீட்டர் ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உடல் வெப்ப நிலை 20.5 GB ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}, {"role": "assistant", "content": "உடல் வெப்ப நிலை இருபது புள்ளி ஐந்து ஜிபி ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை LKR 902.50."}, {"role": "assistant", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை தொள்ளாயிரத்து இரண்டு ரூபாய் ஐம்பது சதம்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை 14வது மாடியில் நடைபெறும்."}, {"role": "assistant", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை பதினான்காவது மாடியில் நடைபெறும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை 26வது மாடியில் நடைபெறும்."}, {"role": "assistant", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை இருபத்தாறாவது மாடியில் நடைபெறும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 256ml இணைய வசதியை Rs. 39.90 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் இருநூற்று ஐம்பத்தாறு மில்லிலிட்டர் இணைய வசதியை முப்பத்தொன்பது ரூபாய் தொண்ணூறு சதம் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 97.7kg இணைய வசதியை ரூ. 74 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் தொண்ணூற்று ஏழு புள்ளி ஏழு கிலோகிராம் இணைய வசதியை எழுபத்து நான்கு ரூபாய் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "மேலதிக விபரங்களுக்கு www.web.de என்ற இணையதளத்தைப் பார்க்கவும்."}, {"role": "assistant", "content": "மேலதிக விபரங்களுக்கு w w w dot web dot de என்ற இணையதளத்தைப் பார்க்கவும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 20.1g இணைய வசதியை Rs. 18 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் இருபது புள்ளி ஒன்று கிராம் இணைய வசதியை பதினெட்டு ரூபாய் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 284 cm இணைய வசதியை ரூ. 46 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் இருநூற்று எண்பத்து நான்கு சென்டிமீட்டர் இணைய வசதியை நாற்பத்தாறு ரூபாய் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "மேலதிக விபரங்களுக்கு https://www.example.com/support என்ற இணையதளத்தைப் பார்க்கவும்."}, {"role": "assistant", "content": "மேலதிக விபரங்களுக்கு h t t p s colon slash slash www dot example dot com slash support என்ற இணையதளத்தைப் பார்க்கவும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை 6வது மாடியில் நடைபெறும்."}, {"role": "assistant", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை ஆறாவது மாடியில் நடைபெறும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 20.3 ml இணைய வசதியை Rs. 757.90 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் இருபது புள்ளி மூன்று மில்லிலிட்டர் இணைய வசதியை எழுநூற்று ஐம்பத்தேழு ரூபாய் தொண்ணூறு சதம் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "மேலதிக விபரங்களுக்கு https://gmail.co.id/faq என்ற இணையதளத்தைப் பார்க்கவும்."}, {"role": "assistant", "content": "மேலதிக விபரங்களுக்கு h t t p s colon slash slash gmail dot co dot id slash faq என்ற இணையதளத்தைப் பார்க்கவும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 34 °C இணைய வசதியை Rs.355 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் முப்பத்து நான்கு டிகிரி செல்சியஸ் இணைய வசதியை முந்நூற்று ஐம்பத்தைந்து ரூபாய் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை 13வது மாடியில் நடைபெறும்."}, {"role": "assistant", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை பதிமூன்றாவது மாடியில் நடைபெறும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை Rs.32,388.50."}, {"role": "assistant", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை முப்பத்திரண்டாயிரத்து முந்நூற்று எண்பத்தெட்டு ரூபாய் ஐம்பது சதம்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை 20வது மாடியில் நடைபெறும்."}, {"role": "assistant", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை இருபதாவது மாடியில் நடைபெறும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 81.4°C இணைய வசதியை Rs. 974.50 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் எண்பத்தொன்று புள்ளி நான்கு டிகிரி செல்சியஸ் இணைய வசதியை தொள்ளாயிரத்து எழுபத்து நான்கு ரூபாய் ஐம்பது சதம் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உடல் வெப்ப நிலை 436kg ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}, {"role": "assistant", "content": "உடல் வெப்ப நிலை நானூற்று முப்பத்தாறு கிலோகிராம் ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "மேலதிக விபரங்களுக்கு https://correo.es/help என்ற இணையதளத்தைப் பார்க்கவும்."}, {"role": "assistant", "content": "மேலதிக விபரங்களுக்கு h t t p s colon slash slash correo dot es slash help என்ற இணையதளத்தைப் பார்க்கவும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "மேலதிக விபரங்களுக்கு https://web.de/help என்ற இணையதளத்தைப் பார்க்கவும்."}, {"role": "assistant", "content": "மேலதிக விபரங்களுக்கு h t t p s colon slash slash web dot de slash help என்ற இணையதளத்தைப் பார்க்கவும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உடல் வெப்ப நிலை 42.5°C ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}, {"role": "assistant", "content": "உடல் வெப்ப நிலை நாற்பத்திரண்டு புள்ளி ஐந்து டிகிரி செல்சியஸ் ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை 11வது மாடியில் நடைபெறும்."}, {"role": "assistant", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை பதினொன்றாவது மாடியில் நடைபெறும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உடல் வெப்ப நிலை 34.1 cm ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}, {"role": "assistant", "content": "உடல் வெப்ப நிலை முப்பத்து நான்கு புள்ளி ஒன்று சென்டிமீட்டர் ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை LKR 49."}, {"role": "assistant", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை நாற்பத்தொன்பது ரூபாய்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை 6ஆவது மாடியில் நடைபெறும்."}, {"role": "assistant", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை ஆறாவது மாடியில் நடைபெறும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை Rs.64.50."}, {"role": "assistant", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை அறுபத்து நான்கு ரூபாய் ஐம்பது சதம்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 211 km இணைய வசதியை Rs. 508.64 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் இருநூற்று பதினொன்று கிலோமீட்டர் இணைய வசதியை ஐந்நூற்று எட்டு ரூபாய் அறுபத்து நான்கு சதம் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உடல் வெப்ப நிலை 208l ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}, {"role": "assistant", "content": "உடல் வெப்ப நிலை இருநூற்று எட்டு லிட்டர் ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உடல் வெப்ப நிலை 97.7kg ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}, {"role": "assistant", "content": "உடல் வெப்ப நிலை தொண்ணூற்று ஏழு புள்ளி ஏழு கிலோகிராம் ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 56.3ml இணைய வசதியை LKR 712 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் ஐம்பத்தாறு புள்ளி மூன்று மில்லிலிட்டர் இணைய வசதியை எழுநூற்று பன்னிரண்டு ரூபாய் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை Rs. 58,507.50."}, {"role": "assistant", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை ஐம்பத்தெட்டாயிரத்து ஐந்நூற்று ஏழு ரூபாய் ஐம்பது சதம்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய டேட்டா பேக் மூலம் 51kg இணைய வசதியை Rs.49,779.50 விலையில் பெற்றுக்கொள்ளுங்கள்."}, {"role": "assistant", "content": "புதிய டேட்டா பேக் மூலம் ஐம்பத்தொன்று கிலோகிராம் இணைய வசதியை நாற்பத்தொன்பதாயிரத்து எழுநூற்று எழுபத்தொன்பது ரூபாய் ஐம்பது சதம் விலையில் பெற்றுக்கொள்ளுங்கள்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உடல் வெப்ப நிலை 100g ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}, {"role": "assistant", "content": "உடல் வெப்ப நிலை நூறு கிராம் ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உடல் வெப்ப நிலை 80GB ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}, {"role": "assistant", "content": "உடல் வெப்ப நிலை எண்பது ஜிபி ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை 12ஆவது மாடியில் நடைபெறும்."}, {"role": "assistant", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை பன்னிரண்டாவது மாடியில் நடைபெறும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "மேலதிக விபரங்களுக்கு www.example.com/login என்ற இணையதளத்தைப் பார்க்கவும்."}, {"role": "assistant", "content": "மேலதிக விபரங்களுக்கு w w w dot example dot com slash login என்ற இணையதளத்தைப் பார்க்கவும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உடல் வெப்ப நிலை 79.8 ml ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}, {"role": "assistant", "content": "உடல் வெப்ப நிலை எழுபத்தொன்பது புள்ளி எட்டு மில்லிலிட்டர் ஆக இருப்பதால் நீங்கள் ஓய்வெடுக்க வேண்டும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை ரூ. 11."}, {"role": "assistant", "content": "தங்கள் கொடுப்பனவு வெற்றிகரமாகப் பெறப்பட்டது, தொகை பதினொன்று ரூபாய்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை 10வது மாடியில் நடைபெறும்."}, {"role": "assistant", "content": "கொழும்பு கிளையில் உங்கள் பரிசோதனை பத்தாவது மாடியில் நடைபெறும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "மேலதிக விபரங்களுக்கு https://example.com என்ற இணையதளத்தைப் பார்க்கவும்."}, {"role": "assistant", "content": "மேலதிக விபரங்களுக்கு h t t p s colon slash slash example dot com என்ற இணையதளத்தைப் பார்க்கவும்."}], "lang": "ta-LK", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 43865 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය හතර තුන අට හය පහ වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 66007 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය හය හය බිංදුව බිංදුව හත වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 62347 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය හය දෙක තුන හතර හත වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 53594 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය පහ තුන පහ නවය හතර ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 5167 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය පහ එක හය හත වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 964302 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය නවය හය හතර තුන බිංදුව දෙක වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 7799 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය හත හත නවය නවය වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 345690 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය තුන හතර පහ හය නවය බිංදුව ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 50218 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය පහ බිංදුව දෙක එක අට ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 437252 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය හතර තුන හත දෙක පහ දෙක වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 48135 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හතර අට එක තුන පහ ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 477843 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය හතර හත හත අට හතර තුන වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 093563 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය බිංදුව නවය තුන පහ හය තුන වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 2090 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය දෙක බිංදුව නවය බිංදුව ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 8081 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය අට බිංදුව අට එක වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 7911 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය හත නවය එක එක වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 4983 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හතර නවය අට තුන ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 63433 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය හය තුන හතර තුන තුන වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 54129 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය පහ හතර එක දෙක නවය වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 28475 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය දෙක අට හතර හත පහ වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 422859 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හතර දෙක දෙක අට පහ නවය ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 7219 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හත දෙක එක නවය ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 39014 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය තුන නවය බිංදුව එක හතර ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 965654 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය නවය හය පහ හය පහ හතර ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 94289 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය නවය හතර දෙක අට නවය වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 6318 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හය තුන එක අට ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 72021 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හත දෙක බිංදුව දෙක එක ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 141634 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය එක හතර එක හය තුන හතර ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 43228 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හතර තුන දෙක දෙක අට ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 83466 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය අට තුන හතර හය හය ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 3825 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය තුන අට දෙක පහ වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 9812 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය නවය අට එක දෙක වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 92564 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය නවය දෙක පහ හය හතර වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 58256 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය පහ අට දෙක පහ හය ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 488441 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය හතර අට අට හතර හතර එක වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 806298 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය අට බිංදුව හය දෙක නවය අට වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 6388 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හය තුන අට අට ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 8466 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය අට හතර හය හය වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 51933 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය පහ එක නවය තුන තුන ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 592250 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය පහ නවය දෙක දෙක පහ බිංදුව වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 7028 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය හත බිංදුව දෙක අට වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 27594 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය දෙක හත පහ නවය හතර වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 8933 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය අට නවය තුන තුන වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 096173 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය බිංදුව නවය හය එක හත තුන වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 62553 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය හය දෙක පහ පහ තුන වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 8544 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය අට පහ හතර හතර වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 108687 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය එක බිංදුව අට හය අට හත ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 2471 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය දෙක හතර හත එක ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 7285 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හත දෙක අට පහ ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 02804 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය බිංදුව දෙක අට බිංදුව හතර ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 8266 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය අට දෙක හය හය වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 73367 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය හත තුන තුන හය හත වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 489559 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හතර අට නවය පහ පහ නවය ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 77232 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හත හත දෙක තුන දෙක ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 5002 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය පහ බිංදුව බිංදුව දෙක වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 4666 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය හතර හය හය හය වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 6676 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හය හය හත හය ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 51475 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය පහ එක හතර හත පහ වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 79717 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය හත නවය හත එක හත වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 50196 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය පහ බිංදුව එක නවය හය වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 6775 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය හය හත හත පහ වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 14687 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය එක හතර හය අට හත ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 676479 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හය හත හය හතර හත නවය ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 312549 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය තුන එක දෙක පහ හතර නවය ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 1396 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය එක තුන නවය හය ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 742178 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය හත හතර දෙක එක හත අට වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 200567 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය දෙක බිංදුව බිංදුව පහ හය හත වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 73684 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය හත තුන හය අට හතර වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 923745 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය නවය දෙක තුන හත හතර පහ වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 61789 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හය එක හත අට නවය ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 651985 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හය පහ එක නවය අට පහ ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 64939 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය හය හතර නවය තුන නවය වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 9719 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය නවය හත එක නවය වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 37759 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය තුන හත හත පහ නවය වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 0112 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය බිංදුව එක එක දෙක වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 6680 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හය හය අට බිංදුව ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 211611 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය දෙක එක එක හය එක එක ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 4768 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය හතර හත හය අට වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 41076 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හතර එක බිංදුව හත හය ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 34278 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය තුන හතර දෙක හත අට වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 04729 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය බිංදුව හතර හත දෙක නවය වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 61550 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හය එක පහ පහ බිංදුව ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 534848 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය පහ තුන හතර අට හතර අට වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 03417 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය බිංදුව තුන හතර එක හත වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 62196 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හය දෙක එක නවය හය ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 55845 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය පහ පහ අට හතර පහ ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 2864 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය දෙක අට හය හතර වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 5003 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය පහ බිංදුව බිංදුව තුන වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 9696 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය නවය හය නවය හය ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 687832 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හය අට හත අට තුන දෙක ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 887372 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය අට අට හත තුන හත දෙක වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 0096 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය බිංදුව බිංදුව නවය හය ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 93930 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය නවය තුන නවය තුන බිංදුව වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 56102 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය පහ හය එක බිංදුව දෙක වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 59729 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය පහ නවය හත දෙක නවය වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 510414 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය පහ එක බිංදුව හතර එක හතර ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 454907 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හතර පහ හතර නවය බිංදුව හත ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 94209 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය නවය හතර දෙක බිංදුව නවය වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 33451 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය තුන තුන හතර පහ එක වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 4968 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හතර නවය හය අට ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 41385 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හතර එක තුන අට පහ ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 69354 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හය නවය තුන පහ හතර ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 35667 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය තුන පහ හය හය හත වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 074833 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය බිංදුව හත හතර අට තුන තුන ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 82929 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය අට දෙක නවය දෙක නවය වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 924134 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය නවය දෙක හතර එක තුන හතර වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 96637 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය නවය හය හය තුන හත වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 373717 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය තුන හත තුන හත එක හත ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 437929 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හතර තුන හත නවය දෙක නවය ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 47264 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය හතර හත දෙක හය හතර වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 7814 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හත අට එක හතර ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 906318 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය නවය බිංදුව හය තුන එක අට වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 531443 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය පහ තුන එක හතර හතර තුන වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 01514 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය බිංදුව එක පහ එක හතර වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 2074 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය දෙක බිංදුව හත හතර ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 405813 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හතර බිංදුව පහ අට එක තුන ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 612038 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය හය එක දෙක බිංදුව තුන අට වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 7990 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය හත නවය නවය බිංදුව වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 08862 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය බිංදුව අට අට හය දෙක වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 09829 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය බිංදුව නවය අට දෙක නවය වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 433933 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හතර තුන තුන නවය තුන තුන ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ සත්‍යාපන කේතය 43679 වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}, {"role": "assistant", "content": "ඔබගේ සත්‍යාපන කේතය හතර තුන හය හත නවය වේ, එය කිසිවෙකු සමඟ බෙදා නොගන්න."}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 576095 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය පහ හත හය බිංදුව නවය පහ ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 4702 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හතර හත බිංදුව දෙක ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 941341 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය නවය හතර එක තුන හතර එක ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 020465 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය බිංදුව දෙක බිංදුව හතර හය පහ ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබගේ ලියාපදිංචි අංකය 44314 ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}, {"role": "assistant", "content": "ඔබගේ ලියාපදිංචි අංකය හතර හතර තුන එක හතර ලෙස නිවැරදිව ඇතුළත් කර ඇත්ද?"}], "lang": "si", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakisend ang 1665 para ma-verify ang iyong identity."}, {"role": "assistant", "content": "Pakisend ang isa anim anim lima para ma-verify ang iyong identity."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ang inyong appointment ay sa 22-30 ng Martes."}, {"role": "assistant", "content": "Ang inyong appointment ay sa dalawampu't dalawa hanggang tatlumpu ng Martes."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ang inyong appointment ay sa 7-11 ng Martes."}, {"role": "assistant", "content": "Ang inyong appointment ay sa pito hanggang labing-isa ng Martes."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paalala: Ang iyong bill ay due na sa 2020."}, {"role": "assistant", "content": "Paalala: Ang iyong bill ay due na sa dalawang libo at dalawampu."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakisend ang 9194 para ma-verify ang iyong identity."}, {"role": "assistant", "content": "Pakisend ang siyam isa siyam apat para ma-verify ang iyong identity."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ang inyong appointment ay sa 6–16 ng Martes."}, {"role": "assistant", "content": "Ang inyong appointment ay sa anim hanggang labing-anim ng Martes."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakikumpirma lang po kung ito ang inyong registered (02) 1321 5741 number."}, {"role": "assistant", "content": "Pakikumpirma lang po kung ito ang inyong registered zero dalawa isa tatlo dalawa isa lima pito apat isa number."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Maaari niyo pong i-download ang form sa www.ziggo.nl."}, {"role": "assistant", "content": "Maaari niyo pong i-download ang form sa w w w dot ziggo dot nl."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakisend ang 620374 para ma-verify ang iyong identity."}, {"role": "assistant", "content": "Pakisend ang anim dalawa zero tatlo pito apat para ma-verify ang iyong identity."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakikumpirma lang po kung ito ang inyong registered (02) 5086 9087 number."}, {"role": "assistant", "content": "Pakikumpirma lang po kung ito ang inyong registered zero dalawa lima zero walo anim siyam zero walo pito number."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakikumpirma lang po kung ito ang inyong registered (02) 4320 8180 number."}, {"role": "assistant", "content": "Pakikumpirma lang po kung ito ang inyong registered zero dalawa apat tatlo dalawa zero walo isa walo zero number."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paalala: Ang iyong bill ay due na sa 1988."}, {"role": "assistant", "content": "Paalala: Ang iyong bill ay due na sa isang libo siyam na raan at walumpu't walo."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakisend ang 93962 para ma-verify ang iyong identity."}, {"role": "assistant", "content": "Pakisend ang siyam tatlo siyam anim dalawa para ma-verify ang iyong identity."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakikumpirma lang po kung ito ang inyong registered +63 978 708 8986 number."}, {"role": "assistant", "content": "Pakikumpirma lang po kung ito ang inyong registered plus anim tatlo siyam pito walo pito zero walo walo siyam walo anim number."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakisend ang 0197 para ma-verify ang iyong identity."}, {"role": "assistant", "content": "Pakisend ang zero isa siyam pito para ma-verify ang iyong identity."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ang inyong appointment ay sa 2-6 ng Martes."}, {"role": "assistant", "content": "Ang inyong appointment ay sa dalawa hanggang anim ng Martes."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ang inyong appointment ay sa 23 - 30 ng Martes."}, {"role": "assistant", "content": "Ang inyong appointment ay sa dalawampu't tatlo hanggang tatlumpu ng Martes."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakikumpirma lang po kung ito ang inyong registered +63 921 999 5184 number."}, {"role": "assistant", "content": "Pakikumpirma lang po kung ito ang inyong registered plus anim tatlo siyam dalawa isa siyam siyam siyam lima isa walo apat number."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paalala: Ang iyong bill ay due na sa 2031."}, {"role": "assistant", "content": "Paalala: Ang iyong bill ay due na sa dalawang libo at tatlumpu't isa."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Maaari niyo pong i-download ang form sa https://outlook.sa."}, {"role": "assistant", "content": "Maaari niyo pong i-download ang form sa outlook dot sa."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ang inyong appointment ay sa 17–25 ng Martes."}, {"role": "assistant", "content": "Ang inyong appointment ay sa labimpito hanggang dalawampu't lima ng Martes."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ang inyong appointment ay sa 5–7 ng Martes."}, {"role": "assistant", "content": "Ang inyong appointment ay sa lima hanggang pito ng Martes."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paalala: Ang iyong bill ay due na sa 1994."}, {"role": "assistant", "content": "Paalala: Ang iyong bill ay due na sa isang libo siyam na raan at siyamnapu't apat."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paalala: Ang iyong bill ay due na sa 1984."}, {"role": "assistant", "content": "Paalala: Ang iyong bill ay due na sa isang libo siyam na raan at walumpu't apat."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Maaari niyo pong i-download ang form sa https://posta.it/faq."}, {"role": "assistant", "content": "Maaari niyo pong i-download ang form sa posta dot it slash faq."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paalala: Ang iyong bill ay due na sa 2017."}, {"role": "assistant", "content": "Paalala: Ang iyong bill ay due na sa dalawang libo at labimpito."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Maaari niyo pong i-download ang form sa https://sapo.pt."}, {"role": "assistant", "content": "Maaari niyo pong i-download ang form sa sapo dot pt."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakikumpirma lang po kung ito ang inyong registered 0915 304 4293 number."}, {"role": "assistant", "content": "Pakikumpirma lang po kung ito ang inyong registered zero siyam isa lima tatlo zero apat apat dalawa siyam tatlo number."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Maaari niyo pong i-download ang form sa https://scicom.com.my."}, {"role": "assistant", "content": "Maaari niyo pong i-download ang form sa scicom dot com dot my."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakisend ang 91014 para ma-verify ang iyong identity."}, {"role": "assistant", "content": "Pakisend ang siyam isa zero isa apat para ma-verify ang iyong identity."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Maaari niyo pong i-download ang form sa www.mail.fr."}, {"role": "assistant", "content": "Maaari niyo pong i-download ang form sa w w w dot mail dot fr."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ang inyong appointment ay sa 14 - 20 ng Martes."}, {"role": "assistant", "content": "Ang inyong appointment ay sa labing-apat hanggang dalawampu ng Martes."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Maaari niyo pong i-download ang form sa www.gmail.com."}, {"role": "assistant", "content": "Maaari niyo pong i-download ang form sa w w w dot gmail dot com."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Maaari niyo pong i-download ang form sa www.mail.fr/login."}, {"role": "assistant", "content": "Maaari niyo pong i-download ang form sa w w w dot mail dot fr slash login."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paalala: Ang iyong bill ay due na sa 2022."}, {"role": "assistant", "content": "Paalala: Ang iyong bill ay due na sa dalawang libo at dalawampu't dalawa."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ang inyong appointment ay sa 14 - 21 ng Martes."}, {"role": "assistant", "content": "Ang inyong appointment ay sa labing-apat hanggang dalawampu't isa ng Martes."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paalala: Ang iyong bill ay due na sa 1999."}, {"role": "assistant", "content": "Paalala: Ang iyong bill ay due na sa isang libo siyam na raan at siyamnapu't siyam."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakikumpirma lang po kung ito ang inyong registered 0980-017-9519 number."}, {"role": "assistant", "content": "Pakikumpirma lang po kung ito ang inyong registered zero siyam walo zero zero isa pito siyam lima isa siyam number."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakikumpirma lang po kung ito ang inyong registered 0903-791-1954 number."}, {"role": "assistant", "content": "Pakikumpirma lang po kung ito ang inyong registered zero siyam zero tatlo pito siyam isa isa siyam lima apat number."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakisend ang 005325 para ma-verify ang iyong identity."}, {"role": "assistant", "content": "Pakisend ang zero zero lima tatlo dalawa lima para ma-verify ang iyong identity."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Maaari niyo pong i-download ang form sa https://sltnet.lk/help."}, {"role": "assistant", "content": "Maaari niyo pong i-download ang form sa sltnet dot lk slash help."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Maaari niyo pong i-download ang form sa www.web.de/login."}, {"role": "assistant", "content": "Maaari niyo pong i-download ang form sa w w w dot web dot de slash login."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Maaari niyo pong i-download ang form sa https://gmail.com."}, {"role": "assistant", "content": "Maaari niyo pong i-download ang form sa gmail dot com."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakikumpirma lang po kung ito ang inyong registered +63 920 407 5064 number."}, {"role": "assistant", "content": "Pakikumpirma lang po kung ito ang inyong registered plus anim tatlo siyam dalawa zero apat zero pito lima zero anim apat number."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Maaari niyo pong i-download ang form sa www.posta.it/login."}, {"role": "assistant", "content": "Maaari niyo pong i-download ang form sa w w w dot posta dot it slash login."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ang inyong appointment ay sa 22–25 ng Martes."}, {"role": "assistant", "content": "Ang inyong appointment ay sa dalawampu't dalawa hanggang dalawampu't lima ng Martes."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paalala: Ang iyong bill ay due na sa 2009."}, {"role": "assistant", "content": "Paalala: Ang iyong bill ay due na sa dalawang libo at siyam."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakisend ang 7480 para ma-verify ang iyong identity."}, {"role": "assistant", "content": "Pakisend ang pito apat walo zero para ma-verify ang iyong identity."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paalala: Ang iyong bill ay due na sa 2018."}, {"role": "assistant", "content": "Paalala: Ang iyong bill ay due na sa dalawang libo at labingwalo."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakikumpirma lang po kung ito ang inyong registered 0976 716 3553 number."}, {"role": "assistant", "content": "Pakikumpirma lang po kung ito ang inyong registered zero siyam pito anim pito isa anim tatlo lima lima tatlo number."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakikumpirma lang po kung ito ang inyong registered 0914 351 3297 number."}, {"role": "assistant", "content": "Pakikumpirma lang po kung ito ang inyong registered zero siyam isa apat tatlo lima isa tatlo dalawa siyam pito number."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakisend ang 153445 para ma-verify ang iyong identity."}, {"role": "assistant", "content": "Pakisend ang isa lima tatlo apat apat lima para ma-verify ang iyong identity."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakisend ang 2852 para ma-verify ang iyong identity."}, {"role": "assistant", "content": "Pakisend ang dalawa walo lima dalawa para ma-verify ang iyong identity."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paalala: Ang iyong bill ay due na sa 1960."}, {"role": "assistant", "content": "Paalala: Ang iyong bill ay due na sa isang libo siyam na raan at animnapu."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ang inyong appointment ay sa 23 - 27 ng Martes."}, {"role": "assistant", "content": "Ang inyong appointment ay sa dalawampu't tatlo hanggang dalawampu't pito ng Martes."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paalala: Ang iyong bill ay due na sa 1954."}, {"role": "assistant", "content": "Paalala: Ang iyong bill ay due na sa isang libo siyam na raan at limampu't apat."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Maaari niyo pong i-download ang form sa https://yahoo.com/faq."}, {"role": "assistant", "content": "Maaari niyo pong i-download ang form sa yahoo dot com slash faq."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Maaari niyo pong i-download ang form sa https://sltnet.lk."}, {"role": "assistant", "content": "Maaari niyo pong i-download ang form sa sltnet dot lk."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakisend ang 31032 para ma-verify ang iyong identity."}, {"role": "assistant", "content": "Pakisend ang tatlo isa zero tatlo dalawa para ma-verify ang iyong identity."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakisend ang 873264 para ma-verify ang iyong identity."}, {"role": "assistant", "content": "Pakisend ang walo pito tatlo dalawa anim apat para ma-verify ang iyong identity."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ang inyong appointment ay sa 20–22 ng Martes."}, {"role": "assistant", "content": "Ang inyong appointment ay sa dalawampu hanggang dalawampu't dalawa ng Martes."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ang inyong appointment ay sa 15-20 ng Martes."}, {"role": "assistant", "content": "Ang inyong appointment ay sa labinlima hanggang dalawampu ng Martes."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakisend ang 75716 para ma-verify ang iyong identity."}, {"role": "assistant", "content": "Pakisend ang pito lima pito isa anim para ma-verify ang iyong identity."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakisend ang 536735 para ma-verify ang iyong identity."}, {"role": "assistant", "content": "Pakisend ang lima tatlo anim pito tatlo lima para ma-verify ang iyong identity."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakikumpirma lang po kung ito ang inyong registered +63 926 476 8540 number."}, {"role": "assistant", "content": "Pakikumpirma lang po kung ito ang inyong registered plus anim tatlo siyam dalawa anim apat pito anim walo lima apat zero number."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paalala: Ang iyong bill ay due na sa 1983."}, {"role": "assistant", "content": "Paalala: Ang iyong bill ay due na sa isang libo siyam na raan at walumpu't tatlo."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Maaari niyo pong i-download ang form sa https://www.gmail.co.id/support."}, {"role": "assistant", "content": "Maaari niyo pong i-download ang form sa w w w dot gmail dot co dot id slash support."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ang inyong appointment ay sa 8-14 ng Martes."}, {"role": "assistant", "content": "Ang inyong appointment ay sa walo hanggang labing-apat ng Martes."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakikumpirma lang po kung ito ang inyong registered 0948-980-9852 number."}, {"role": "assistant", "content": "Pakikumpirma lang po kung ito ang inyong registered zero siyam apat walo siyam walo zero siyam walo lima dalawa number."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakikumpirma lang po kung ito ang inyong registered 0999 259 5095 number."}, {"role": "assistant", "content": "Pakikumpirma lang po kung ito ang inyong registered zero siyam siyam siyam dalawa lima siyam lima zero siyam lima number."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakikumpirma lang po kung ito ang inyong registered 0961 183 6663 number."}, {"role": "assistant", "content": "Pakikumpirma lang po kung ito ang inyong registered zero siyam anim isa isa walo tatlo anim anim anim tatlo number."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakikumpirma lang po kung ito ang inyong registered +63 965 036 6365 number."}, {"role": "assistant", "content": "Pakikumpirma lang po kung ito ang inyong registered plus anim tatlo siyam anim lima zero tatlo anim anim tatlo anim lima number."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakisend ang 79845 para ma-verify ang iyong identity."}, {"role": "assistant", "content": "Pakisend ang pito siyam walo apat lima para ma-verify ang iyong identity."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakikumpirma lang po kung ito ang inyong registered (02) 9157 9012 number."}, {"role": "assistant", "content": "Pakikumpirma lang po kung ito ang inyong registered zero dalawa siyam isa lima pito siyam zero isa dalawa number."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakisend ang 66554 para ma-verify ang iyong identity."}, {"role": "assistant", "content": "Pakisend ang anim anim lima lima apat para ma-verify ang iyong identity."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paalala: Ang iyong bill ay due na sa 1979."}, {"role": "assistant", "content": "Paalala: Ang iyong bill ay due na sa isang libo siyam na raan at pitumpu't siyam."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ang inyong appointment ay sa 21-25 ng Martes."}, {"role": "assistant", "content": "Ang inyong appointment ay sa dalawampu't isa hanggang dalawampu't lima ng Martes."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ang inyong appointment ay sa 7 - 12 ng Martes."}, {"role": "assistant", "content": "Ang inyong appointment ay sa pito hanggang labindalawa ng Martes."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakikumpirma lang po kung ito ang inyong registered (02) 8007 2211 number."}, {"role": "assistant", "content": "Pakikumpirma lang po kung ito ang inyong registered zero dalawa walo zero zero pito dalawa dalawa isa isa number."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Maaari niyo pong i-download ang form sa https://company.co/help."}, {"role": "assistant", "content": "Maaari niyo pong i-download ang form sa company dot co slash help."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakisend ang 287119 para ma-verify ang iyong identity."}, {"role": "assistant", "content": "Pakisend ang dalawa walo pito isa isa siyam para ma-verify ang iyong identity."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paalala: Ang iyong bill ay due na sa 2003."}, {"role": "assistant", "content": "Paalala: Ang iyong bill ay due na sa dalawang libo at tatlo."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakisend ang 919215 para ma-verify ang iyong identity."}, {"role": "assistant", "content": "Pakisend ang siyam isa siyam dalawa isa lima para ma-verify ang iyong identity."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Maaari niyo pong i-download ang form sa https://www.gmail.com/support."}, {"role": "assistant", "content": "Maaari niyo pong i-download ang form sa w w w dot gmail dot com slash support."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakikumpirma lang po kung ito ang inyong registered +63 928 066 6700 number."}, {"role": "assistant", "content": "Pakikumpirma lang po kung ito ang inyong registered plus anim tatlo siyam dalawa walo zero anim anim anim pito zero zero number."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakikumpirma lang po kung ito ang inyong registered (02) 2956 4542 number."}, {"role": "assistant", "content": "Pakikumpirma lang po kung ito ang inyong registered zero dalawa dalawa siyam lima anim apat lima apat dalawa number."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakikumpirma lang po kung ito ang inyong registered +63 954 253 5960 number."}, {"role": "assistant", "content": "Pakikumpirma lang po kung ito ang inyong registered plus anim tatlo siyam lima apat dalawa lima tatlo lima siyam anim zero number."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ang inyong appointment ay sa 4–6 ng Martes."}, {"role": "assistant", "content": "Ang inyong appointment ay sa apat hanggang anim ng Martes."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ang inyong appointment ay sa 2 - 11 ng Martes."}, {"role": "assistant", "content": "Ang inyong appointment ay sa dalawa hanggang labing-isa ng Martes."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakikumpirma lang po kung ito ang inyong registered 0983-525-4279 number."}, {"role": "assistant", "content": "Pakikumpirma lang po kung ito ang inyong registered zero siyam walo tatlo lima dalawa lima apat dalawa pito siyam number."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakikumpirma lang po kung ito ang inyong registered +63 922 306 4362 number."}, {"role": "assistant", "content": "Pakikumpirma lang po kung ito ang inyong registered plus anim tatlo siyam dalawa dalawa tatlo zero anim apat tatlo anim dalawa number."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakisend ang 764539 para ma-verify ang iyong identity."}, {"role": "assistant", "content": "Pakisend ang pito anim apat lima tatlo siyam para ma-verify ang iyong identity."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ang inyong appointment ay sa 7 - 9 ng Martes."}, {"role": "assistant", "content": "Ang inyong appointment ay sa pito hanggang siyam ng Martes."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ang inyong appointment ay sa 11-20 ng Martes."}, {"role": "assistant", "content": "Ang inyong appointment ay sa labing-isa hanggang dalawampu ng Martes."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakisend ang 990457 para ma-verify ang iyong identity."}, {"role": "assistant", "content": "Pakisend ang siyam siyam zero apat lima pito para ma-verify ang iyong identity."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Maaari niyo pong i-download ang form sa https://www.sltnet.lk/support."}, {"role": "assistant", "content": "Maaari niyo pong i-download ang form sa w w w dot sltnet dot lk slash support."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paalala: Ang iyong bill ay due na sa 1957."}, {"role": "assistant", "content": "Paalala: Ang iyong bill ay due na sa isang libo siyam na raan at limampu't pito."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Maaari niyo pong i-download ang form sa https://web.de/help."}, {"role": "assistant", "content": "Maaari niyo pong i-download ang form sa web dot de slash help."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Maaari niyo pong i-download ang form sa https://mail.fr/help."}, {"role": "assistant", "content": "Maaari niyo pong i-download ang form sa mail dot fr slash help."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ang inyong appointment ay sa 28–34 ng Martes."}, {"role": "assistant", "content": "Ang inyong appointment ay sa dalawampu't walo hanggang tatlumpu't apat ng Martes."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakisend ang 3332 para ma-verify ang iyong identity."}, {"role": "assistant", "content": "Pakisend ang tatlo tatlo tatlo dalawa para ma-verify ang iyong identity."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakikumpirma lang po kung ito ang inyong registered (02) 9249 6611 number."}, {"role": "assistant", "content": "Pakikumpirma lang po kung ito ang inyong registered zero dalawa siyam dalawa apat siyam anim anim isa isa number."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Maaari niyo pong i-download ang form sa https://mail.fr/faq."}, {"role": "assistant", "content": "Maaari niyo pong i-download ang form sa mail dot fr slash faq."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakikumpirma lang po kung ito ang inyong registered 0929 761 2197 number."}, {"role": "assistant", "content": "Pakikumpirma lang po kung ito ang inyong registered zero siyam dalawa siyam pito anim isa dalawa isa siyam pito number."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Maaari niyo pong i-download ang form sa https://www.yahoo.com.ph/support."}, {"role": "assistant", "content": "Maaari niyo pong i-download ang form sa w w w dot yahoo dot com dot ph slash support."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakisend ang 8606 para ma-verify ang iyong identity."}, {"role": "assistant", "content": "Pakisend ang walo anim zero anim para ma-verify ang iyong identity."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paalala: Ang iyong bill ay due na sa 2032."}, {"role": "assistant", "content": "Paalala: Ang iyong bill ay due na sa dalawang libo at tatlumpu't dalawa."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakikumpirma lang po kung ito ang inyong registered (02) 4042 6352 number."}, {"role": "assistant", "content": "Pakikumpirma lang po kung ito ang inyong registered zero dalawa apat zero apat dalawa anim tatlo lima dalawa number."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ang inyong appointment ay sa 28-36 ng Martes."}, {"role": "assistant", "content": "Ang inyong appointment ay sa dalawampu't walo hanggang tatlumpu't anim ng Martes."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakisend ang 086488 para ma-verify ang iyong identity."}, {"role": "assistant", "content": "Pakisend ang zero walo anim apat walo walo para ma-verify ang iyong identity."}], "lang": "tl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "للاستفسار عن حالة طلبك X-12340567، يرجى الاتصال بنا على 0546216018."}, {"role": "assistant", "content": "للاستفسار عن حالة طلبك X واحد اثنان ثلاثة أربعة صفر خمسة ستة سبعة، يرجى الاتصال بنا على صفر خمسة أربعة ستة اثنان واحد ستة صفر واحد ثمانية."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "سيصلك المندوب خلال 4 دقائق، يرجى البقاء بالقرب من 0561219577."}, {"role": "assistant", "content": "سيصلك المندوب خلال أربعة دقائق، يرجى البقاء بالقرب من صفر خمسة ستة واحد اثنان واحد تسعة خمسة سبعة سبعة."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "سيصلك المندوب خلال ٦٣٥ دقائق، يرجى البقاء بالقرب من +٩٦٦ ٥٨ ٥٥٨ ٥٦٦٢."}, {"role": "assistant", "content": "سيصلك المندوب خلال ستمائة وخمسة وثلاثون دقائق، يرجى البقاء بالقرب من زائد تسعة ستة ستة خمسة ثمانية خمسة خمسة ثمانية خمسة ستة ستة اثنان."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "أرسل النموذج إلى chen8875@company.co أو تفضل بزيارة www.yahoo.com/login."}, {"role": "assistant", "content": "أرسل النموذج إلى chen ثمانية ثمانية سبعة خمسة آت company نقطة co أو تفضل بزيارة w w w نقطة yahoo نقطة com شرطة مائلة login."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "سيصلك المندوب خلال 107 دقائق، يرجى البقاء بالقرب من 051 967 1952."}, {"role": "assistant", "content": "سيصلك المندوب خلال مائة وسبعة دقائق، يرجى البقاء بالقرب من صفر خمسة واحد تسعة ستة سبعة واحد تسعة خمسة اثنان."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تكلفة الكشف الطبي هي EGP 56.50، يمكنكم الدفع عند الوصول"}, {"role": "assistant", "content": "تكلفة الكشف الطبي هي ستة وخمسون جنيهاً وخمسون قرشاً، يمكنكم الدفع عند الوصول"}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تكلفة الكشف الطبي هي SAR 445.50، يمكنكم الدفع عند الوصول"}, {"role": "assistant", "content": "تكلفة الكشف الطبي هي أربعمائة وخمسة وأربعون ريالاً وخمسون هللةً، يمكنكم الدفع عند الوصول"}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "للاستفسار عن حالة طلبك PO 55219، يرجى الاتصال بنا على 0503909391."}, {"role": "assistant", "content": "للاستفسار عن حالة طلبك PO خمسة خمسة اثنان واحد تسعة، يرجى الاتصال بنا على صفر خمسة صفر ثلاثة تسعة صفر تسعة ثلاثة تسعة واحد."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا https://correo.es/help."}, {"role": "assistant", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا correo نقطة es شرطة مائلة help."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "سيصلك المندوب خلال ٩٥ دقائق، يرجى البقاء بالقرب من ٠٥٥ ٥٢١ ٢٩٧٣."}, {"role": "assistant", "content": "سيصلك المندوب خلال خمسة وتسعون دقائق، يرجى البقاء بالقرب من صفر خمسة خمسة خمسة اثنان واحد اثنان تسعة سبعة ثلاثة."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا https://www.example.com/support."}, {"role": "assistant", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا w w w نقطة example نقطة com شرطة مائلة support."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا https://onet.pl."}, {"role": "assistant", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا onet نقطة pl."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "للاستفسار عن حالة طلبك TK-88213، يرجى الاتصال بنا على 0568427015."}, {"role": "assistant", "content": "للاستفسار عن حالة طلبك TK ثمانية ثمانية اثنان واحد ثلاثة، يرجى الاتصال بنا على صفر خمسة ستة ثمانية أربعة اثنان سبعة صفر واحد خمسة."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "سيصلك المندوب خلال 8 دقائق، يرجى البقاء بالقرب من 0598153718."}, {"role": "assistant", "content": "سيصلك المندوب خلال ثمانية دقائق، يرجى البقاء بالقرب من صفر خمسة تسعة ثمانية واحد خمسة ثلاثة سبعة واحد ثمانية."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تكلفة الكشف الطبي هي USD ٣٤١، يمكنكم الدفع عند الوصول"}, {"role": "assistant", "content": "تكلفة الكشف الطبي هي ثلاثمائة وواحد وأربعون دولاراً، يمكنكم الدفع عند الوصول"}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تكلفة الكشف الطبي هي $35.50، يمكنكم الدفع عند الوصول"}, {"role": "assistant", "content": "تكلفة الكشف الطبي هي خمسة وثلاثون دولاراً وخمسون سنتاً، يمكنكم الدفع عند الوصول"}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تكلفة الكشف الطبي هي USD ٣٨٦.٥٠، يمكنكم الدفع عند الوصول"}, {"role": "assistant", "content": "تكلفة الكشف الطبي هي ثلاثمائة وستة وثمانون دولاراً وخمسون سنتاً، يمكنكم الدفع عند الوصول"}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا www.sltnet.lk."}, {"role": "assistant", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا w w w نقطة sltnet نقطة lk."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا www.outlook.sa."}, {"role": "assistant", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا w w w نقطة outlook نقطة sa."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "سيصلك المندوب خلال 350 دقائق، يرجى البقاء بالقرب من +966 56 695 7310."}, {"role": "assistant", "content": "سيصلك المندوب خلال ثلاثمائة وخمسون دقائق، يرجى البقاء بالقرب من زائد تسعة ستة ستة خمسة ستة ستة تسعة خمسة سبعة ثلاثة واحد صفر."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا www.correo.es/login."}, {"role": "assistant", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا w w w نقطة correo نقطة es شرطة مائلة login."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا https://mail.fr/faq."}, {"role": "assistant", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا mail نقطة fr شرطة مائلة faq."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "للاستفسار عن حالة طلبك ABC1234، يرجى الاتصال بنا على +966 51 945 2024."}, {"role": "assistant", "content": "للاستفسار عن حالة طلبك ABC واحد اثنان ثلاثة أربعة، يرجى الاتصال بنا على زائد تسعة ستة ستة خمسة واحد تسعة أربعة خمسة اثنان صفر اثنان أربعة."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تكلفة الكشف الطبي هي $528.90، يمكنكم الدفع عند الوصول"}, {"role": "assistant", "content": "تكلفة الكشف الطبي هي خمسمائة وثمانية وعشرون دولاراً وتسعون سنتاً، يمكنكم الدفع عند الوصول"}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "سيصلك المندوب خلال ٦ دقائق، يرجى البقاء بالقرب من ٠٥٧٦٨٨٠١١٦."}, {"role": "assistant", "content": "سيصلك المندوب خلال ستة دقائق، يرجى البقاء بالقرب من صفر خمسة سبعة ستة ثمانية ثمانية صفر واحد واحد ستة."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تكلفة الكشف الطبي هي EGP 245، يمكنكم الدفع عند الوصول"}, {"role": "assistant", "content": "تكلفة الكشف الطبي هي مئتان وخمسة وأربعون جنيهاً، يمكنكم الدفع عند الوصول"}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "سيصلك المندوب خلال ٦٧ دقائق، يرجى البقاء بالقرب من ٠٥٤٦٦٢١٧٤٦."}, {"role": "assistant", "content": "سيصلك المندوب خلال سبعة وستون دقائق، يرجى البقاء بالقرب من صفر خمسة أربعة ستة ستة اثنان واحد سبعة أربعة ستة."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "سيصلك المندوب خلال 50 دقائق، يرجى البقاء بالقرب من +966 56 070 8159."}, {"role": "assistant", "content": "سيصلك المندوب خلال خمسون دقائق، يرجى البقاء بالقرب من زائد تسعة ستة ستة خمسة ستة صفر سبعة صفر ثمانية واحد خمسة تسعة."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "للاستفسار عن حالة طلبك WXY 1234، يرجى الاتصال بنا على 010 2329 7983."}, {"role": "assistant", "content": "للاستفسار عن حالة طلبك WXY واحد اثنان ثلاثة أربعة، يرجى الاتصال بنا على صفر واحد صفر اثنان ثلاثة اثنان تسعة سبعة تسعة ثمانية ثلاثة."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "أرسل النموذج إلى nurul_h٨٤@ziggo.nl أو تفضل بزيارة https://gmail.com/faq."}, {"role": "assistant", "content": "أرسل النموذج إلى nurul underscore h ثمانية أربعة آت ziggo نقطة nl أو تفضل بزيارة gmail نقطة com شرطة مائلة faq."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تكلفة الكشف الطبي هي EGP ٢٩.٥٢، يمكنكم الدفع عند الوصول"}, {"role": "assistant", "content": "تكلفة الكشف الطبي هي تسعة وعشرون جنيهاً واثنان وخمسون قرشاً، يمكنكم الدفع عند الوصول"}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "للاستفسار عن حالة طلبك ABC1234، يرجى الاتصال بنا على 053 661 7719."}, {"role": "assistant", "content": "للاستفسار عن حالة طلبك ABC واحد اثنان ثلاثة أربعة، يرجى الاتصال بنا على صفر خمسة ثلاثة ستة ستة واحد سبعة سبعة واحد تسعة."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "أرسل النموذج إلى juan@sapo.pt أو تفضل بزيارة https://www.posta.it/support."}, {"role": "assistant", "content": "أرسل النموذج إلى juan آت sapo نقطة pt أو تفضل بزيارة w w w نقطة posta نقطة it شرطة مائلة support."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "أرسل النموذج إلى jan.devries@gmail.co.id أو تفضل بزيارة https://gmail.com/faq."}, {"role": "assistant", "content": "أرسل النموذج إلى jan نقطة devries آت gmail نقطة co نقطة id أو تفضل بزيارة gmail نقطة com شرطة مائلة faq."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "أرسل النموذج إلى ravi22@onet.pl أو تفضل بزيارة https://yahoo.com/faq."}, {"role": "assistant", "content": "أرسل النموذج إلى ravi اثنان اثنان آت onet نقطة pl أو تفضل بزيارة yahoo نقطة com شرطة مائلة faq."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا www.gmail.co.id/login."}, {"role": "assistant", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا w w w نقطة gmail نقطة co نقطة id شرطة مائلة login."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا www.yahoo.com.ph."}, {"role": "assistant", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا w w w نقطة yahoo نقطة com نقطة ph."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "أرسل النموذج إلى billing@outlook.sa أو تفضل بزيارة www.bank.my."}, {"role": "assistant", "content": "أرسل النموذج إلى billing آت outlook نقطة sa أو تفضل بزيارة w w w نقطة bank نقطة my."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تكلفة الكشف الطبي هي 10,177.90 ج.م، يمكنكم الدفع عند الوصول"}, {"role": "assistant", "content": "تكلفة الكشف الطبي هي عشرة آلاف ومائة وسبعة وسبعون جنيهاً وتسعون قرشاً، يمكنكم الدفع عند الوصول"}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا https://www.yahoo.com.ph/support."}, {"role": "assistant", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا w w w نقطة yahoo نقطة com نقطة ph شرطة مائلة support."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "للاستفسار عن حالة طلبك WXY 1234، يرجى الاتصال بنا على +966 54 412 0141."}, {"role": "assistant", "content": "للاستفسار عن حالة طلبك WXY واحد اثنان ثلاثة أربعة، يرجى الاتصال بنا على زائد تسعة ستة ستة خمسة أربعة أربعة واحد اثنان صفر واحد أربعة واحد."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "سيصلك المندوب خلال 6 دقائق، يرجى البقاء بالقرب من 013 5817 4972."}, {"role": "assistant", "content": "سيصلك المندوب خلال ستة دقائق، يرجى البقاء بالقرب من صفر واحد ثلاثة خمسة ثمانية واحد سبعة أربعة تسعة سبعة اثنان."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا https://outlook.sa/help."}, {"role": "assistant", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا outlook نقطة sa شرطة مائلة help."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "سيصلك المندوب خلال 2 دقائق، يرجى البقاء بالقرب من +966 56 634 7618."}, {"role": "assistant", "content": "سيصلك المندوب خلال اثنان دقائق، يرجى البقاء بالقرب من زائد تسعة ستة ستة خمسة ستة ستة ثلاثة أربعة سبعة ستة واحد ثمانية."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "للاستفسار عن حالة طلبك INV0472، يرجى الاتصال بنا على 0574309783."}, {"role": "assistant", "content": "للاستفسار عن حالة طلبك INV صفر أربعة سبعة اثنان، يرجى الاتصال بنا على صفر خمسة سبعة أربعة ثلاثة صفر تسعة سبعة ثمانية ثلاثة."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "أرسل النموذج إلى nurul_h@web.de أو تفضل بزيارة https://sltnet.lk/faq."}, {"role": "assistant", "content": "أرسل النموذج إلى nurul underscore h آت web نقطة de أو تفضل بزيارة sltnet نقطة lk شرطة مائلة faq."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تكلفة الكشف الطبي هي AED 27، يمكنكم الدفع عند الوصول"}, {"role": "assistant", "content": "تكلفة الكشف الطبي هي سبعة وعشرون درهماً، يمكنكم الدفع عند الوصول"}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "أرسل النموذج إلى kumar@gmail.com أو تفضل بزيارة https://sapo.pt/faq."}, {"role": "assistant", "content": "أرسل النموذج إلى kumar آت gmail نقطة com أو تفضل بزيارة sapo نقطة pt شرطة مائلة faq."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "أرسل النموذج إلى jean.dupont@ziggo.nl أو تفضل بزيارة https://scicom.com.my."}, {"role": "assistant", "content": "أرسل النموذج إلى jean نقطة dupont آت ziggo نقطة nl أو تفضل بزيارة scicom نقطة com نقطة my."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "سيصلك المندوب خلال 25 دقائق، يرجى البقاء بالقرب من 0585066535."}, {"role": "assistant", "content": "سيصلك المندوب خلال خمسة وعشرون دقائق، يرجى البقاء بالقرب من صفر خمسة ثمانية خمسة صفر ستة ستة خمسة ثلاثة خمسة."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تكلفة الكشف الطبي هي AED 25، يمكنكم الدفع عند الوصول"}, {"role": "assistant", "content": "تكلفة الكشف الطبي هي خمسة وعشرون درهماً، يمكنكم الدفع عند الوصول"}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "للاستفسار عن حالة طلبك REF-٢٠٢٤-٠١١٣، يرجى الاتصال بنا على ٠٥٢٤١٩٨٦٦٧."}, {"role": "assistant", "content": "للاستفسار عن حالة طلبك REF اثنان صفر اثنان أربعة صفر واحد واحد ثلاثة، يرجى الاتصال بنا على صفر خمسة اثنان أربعة واحد تسعة ثمانية ستة ستة سبعة."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا www.correo.es."}, {"role": "assistant", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا w w w نقطة correo نقطة es."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "للاستفسار عن حالة طلبك B2-14، يرجى الاتصال بنا على 055 657 6847."}, {"role": "assistant", "content": "للاستفسار عن حالة طلبك B اثنان واحد أربعة، يرجى الاتصال بنا على صفر خمسة خمسة ستة خمسة سبعة ستة ثمانية أربعة سبعة."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا https://onet.pl/faq."}, {"role": "assistant", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا onet نقطة pl شرطة مائلة faq."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "أرسل النموذج إلى ali53@company.co أو تفضل بزيارة https://yahoo.com/faq."}, {"role": "assistant", "content": "أرسل النموذج إلى ali خمسة ثلاثة آت company نقطة co أو تفضل بزيارة yahoo نقطة com شرطة مائلة faq."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تكلفة الكشف الطبي هي ٨٥ دولار، يمكنكم الدفع عند الوصول"}, {"role": "assistant", "content": "تكلفة الكشف الطبي هي خمسة وثمانون دولاراً، يمكنكم الدفع عند الوصول"}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "أرسل النموذج إلى luca90@gmail.com أو تفضل بزيارة https://www.gmail.co.id/support."}, {"role": "assistant", "content": "أرسل النموذج إلى luca تسعة صفر آت gmail نقطة com أو تفضل بزيارة w w w نقطة gmail نقطة co نقطة id شرطة مائلة support."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "أرسل النموذج إلى mei.ling81@gmail.co.id أو تفضل بزيارة www.sapo.pt/login."}, {"role": "assistant", "content": "أرسل النموذج إلى mei نقطة ling ثمانية واحد آت gmail نقطة co نقطة id أو تفضل بزيارة w w w نقطة sapo نقطة pt شرطة مائلة login."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "للاستفسار عن حالة طلبك ABC1234، يرجى الاتصال بنا على 012 0155 4733."}, {"role": "assistant", "content": "للاستفسار عن حالة طلبك ABC واحد اثنان ثلاثة أربعة، يرجى الاتصال بنا على صفر واحد اثنان صفر واحد خمسة خمسة أربعة سبعة ثلاثة ثلاثة."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا https://www.mail.fr/support."}, {"role": "assistant", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا w w w نقطة mail نقطة fr شرطة مائلة support."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تكلفة الكشف الطبي هي EGP 134.27، يمكنكم الدفع عند الوصول"}, {"role": "assistant", "content": "تكلفة الكشف الطبي هي مائة وأربعة وثلاثون جنيهاً وسبعة وعشرون قرشاً، يمكنكم الدفع عند الوصول"}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا https://www.correo.es/support."}, {"role": "assistant", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا w w w نقطة correo نقطة es شرطة مائلة support."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تكلفة الكشف الطبي هي EGP 938.50، يمكنكم الدفع عند الوصول"}, {"role": "assistant", "content": "تكلفة الكشف الطبي هي تسعمائة وثمانية وثلاثون جنيهاً وخمسون قرشاً، يمكنكم الدفع عند الوصول"}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "للاستفسار عن حالة طلبك X-12340567، يرجى الاتصال بنا على +966 54 949 4963."}, {"role": "assistant", "content": "للاستفسار عن حالة طلبك X واحد اثنان ثلاثة أربعة صفر خمسة ستة سبعة، يرجى الاتصال بنا على زائد تسعة ستة ستة خمسة أربعة تسعة أربعة تسعة أربعة تسعة ستة ثلاثة."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "سيصلك المندوب خلال 85 دقائق، يرجى البقاء بالقرب من 053 610 8941."}, {"role": "assistant", "content": "سيصلك المندوب خلال خمسة وثمانون دقائق، يرجى البقاء بالقرب من صفر خمسة ثلاثة ستة واحد صفر ثمانية تسعة أربعة واحد."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا www.yahoo.com."}, {"role": "assistant", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا w w w نقطة yahoo نقطة com."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "أرسل النموذج إلى juan@yahoo.com.ph أو تفضل بزيارة https://web.de/faq."}, {"role": "assistant", "content": "أرسل النموذج إلى juan آت yahoo نقطة com نقطة ph أو تفضل بزيارة web نقطة de شرطة مائلة faq."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تكلفة الكشف الطبي هي ٨٧٣ جنيه، يمكنكم الدفع عند الوصول"}, {"role": "assistant", "content": "تكلفة الكشف الطبي هي ثمانمائة وثلاثة وسبعون جنيهاً، يمكنكم الدفع عند الوصول"}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تكلفة الكشف الطبي هي 4,098.50 ر.س، يمكنكم الدفع عند الوصول"}, {"role": "assistant", "content": "تكلفة الكشف الطبي هي أربعة آلاف وثمانية وتسعون ريالاً وخمسون هللةً، يمكنكم الدفع عند الوصول"}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "للاستفسار عن حالة طلبك X-12340567، يرجى الاتصال بنا على 0506538120."}, {"role": "assistant", "content": "للاستفسار عن حالة طلبك X واحد اثنان ثلاثة أربعة صفر خمسة ستة سبعة، يرجى الاتصال بنا على صفر خمسة صفر ستة خمسة ثلاثة ثمانية واحد اثنان صفر."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "سيصلك المندوب خلال 846 دقائق، يرجى البقاء بالقرب من 0586070542."}, {"role": "assistant", "content": "سيصلك المندوب خلال ثمانمائة وستة وأربعون دقائق، يرجى البقاء بالقرب من صفر خمسة ثمانية ستة صفر سبعة صفر خمسة أربعة اثنان."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "للاستفسار عن حالة طلبك PO 55219، يرجى الاتصال بنا على +966 59 380 9873."}, {"role": "assistant", "content": "للاستفسار عن حالة طلبك PO خمسة خمسة اثنان واحد تسعة، يرجى الاتصال بنا على زائد تسعة ستة ستة خمسة تسعة ثلاثة ثمانية صفر تسعة ثمانية سبعة ثلاثة."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "للاستفسار عن حالة طلبك ABC١٢٣٤، يرجى الاتصال بنا على +٩٦٦ ٥٩ ٠٥٣ ٠٩٢٩."}, {"role": "assistant", "content": "للاستفسار عن حالة طلبك ABC واحد اثنان ثلاثة أربعة، يرجى الاتصال بنا على زائد تسعة ستة ستة خمسة تسعة صفر خمسة ثلاثة صفر تسعة اثنان تسعة."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا https://company.co/faq."}, {"role": "assistant", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا company نقطة co شرطة مائلة faq."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "سيصلك المندوب خلال 1 دقائق، يرجى البقاء بالقرب من 017 4771 9816."}, {"role": "assistant", "content": "سيصلك المندوب خلال واحد دقائق، يرجى البقاء بالقرب من صفر واحد سبعة أربعة سبعة سبعة واحد تسعة ثمانية واحد ستة."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "سيصلك المندوب خلال 927 دقائق، يرجى البقاء بالقرب من 053 909 3021."}, {"role": "assistant", "content": "سيصلك المندوب خلال تسعمائة وسبعة وعشرون دقائق، يرجى البقاء بالقرب من صفر خمسة ثلاثة تسعة صفر تسعة ثلاثة صفر اثنان واحد."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تكلفة الكشف الطبي هي 7,961.91 ج.م، يمكنكم الدفع عند الوصول"}, {"role": "assistant", "content": "تكلفة الكشف الطبي هي سبعة آلاف وتسعمائة وواحد وستون جنيهاً وواحد وتسعون قرشاً، يمكنكم الدفع عند الوصول"}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "للاستفسار عن حالة طلبك B2-14، يرجى الاتصال بنا على 018 7681 7545."}, {"role": "assistant", "content": "للاستفسار عن حالة طلبك B اثنان واحد أربعة، يرجى الاتصال بنا على صفر واحد ثمانية سبعة ستة ثمانية واحد سبعة خمسة أربعة خمسة."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "سيصلك المندوب خلال 9 دقائق، يرجى البقاء بالقرب من 012 1503 9512."}, {"role": "assistant", "content": "سيصلك المندوب خلال تسعة دقائق، يرجى البقاء بالقرب من صفر واحد اثنان واحد خمسة صفر ثلاثة تسعة خمسة واحد اثنان."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "سيصلك المندوب خلال 604 دقائق، يرجى البقاء بالقرب من +966 50 062 1072."}, {"role": "assistant", "content": "سيصلك المندوب خلال ستمائة وأربعة دقائق، يرجى البقاء بالقرب من زائد تسعة ستة ستة خمسة صفر صفر ستة اثنان واحد صفر سبعة اثنان."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تكلفة الكشف الطبي هي ٢٠ ريال، يمكنكم الدفع عند الوصول"}, {"role": "assistant", "content": "تكلفة الكشف الطبي هي عشرون ريالاً، يمكنكم الدفع عند الوصول"}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "للاستفسار عن حالة طلبك X-١٢٣٤٠٥٦٧، يرجى الاتصال بنا على +٩٦٦ ٥٢ ٤٧٣ ٥٨٧١."}, {"role": "assistant", "content": "للاستفسار عن حالة طلبك X واحد اثنان ثلاثة أربعة صفر خمسة ستة سبعة، يرجى الاتصال بنا على زائد تسعة ستة ستة خمسة اثنان أربعة سبعة ثلاثة خمسة ثمانية سبعة واحد."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "سيصلك المندوب خلال ٦٠٥ دقائق، يرجى البقاء بالقرب من ٠٥٢٦٤٨٥٨٩٤."}, {"role": "assistant", "content": "سيصلك المندوب خلال ستمائة وخمسة دقائق، يرجى البقاء بالقرب من صفر خمسة اثنان ستة أربعة ثمانية خمسة ثمانية تسعة أربعة."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا https://bank.my/help."}, {"role": "assistant", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا bank نقطة my شرطة مائلة help."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "أرسل النموذج إلى help@posta.it أو تفضل بزيارة www.web.de."}, {"role": "assistant", "content": "أرسل النموذج إلى help آت posta نقطة it أو تفضل بزيارة w w w نقطة web نقطة de."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "للاستفسار عن حالة طلبك ABC١٢٣٤، يرجى الاتصال بنا على ٠٥٨ ٠٢٦ ١٢٥٥."}, {"role": "assistant", "content": "للاستفسار عن حالة طلبك ABC واحد اثنان ثلاثة أربعة، يرجى الاتصال بنا على صفر خمسة ثمانية صفر اثنان ستة واحد اثنان خمسة خمسة."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "أرسل النموذج إلى billing@gmail.co.id أو تفضل بزيارة https://scicom.com.my/help."}, {"role": "assistant", "content": "أرسل النموذج إلى billing آت gmail نقطة co نقطة id أو تفضل بزيارة scicom نقطة com نقطة my شرطة مائلة help."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "أرسل النموذج إلى chen88@yahoo.com أو تفضل بزيارة www.bank.my."}, {"role": "assistant", "content": "أرسل النموذج إلى chen ثمانية ثمانية آت yahoo نقطة com أو تفضل بزيارة w w w نقطة bank نقطة my."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "أرسل النموذج إلى fatima@posta.it أو تفضل بزيارة https://www.bank.my/support."}, {"role": "assistant", "content": "أرسل النموذج إلى fatima آت posta نقطة it أو تفضل بزيارة w w w نقطة bank نقطة my شرطة مائلة support."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "سيصلك المندوب خلال ٨٢ دقائق، يرجى البقاء بالقرب من ٠٥٩ ٨٠٦ ٦٩٨٤."}, {"role": "assistant", "content": "سيصلك المندوب خلال اثنان وثمانون دقائق، يرجى البقاء بالقرب من صفر خمسة تسعة ثمانية صفر ستة ستة تسعة ثمانية أربعة."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "أرسل النموذج إلى siti@web.de أو تفضل بزيارة https://posta.it."}, {"role": "assistant", "content": "أرسل النموذج إلى siti آت web نقطة de أو تفضل بزيارة posta نقطة it."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تكلفة الكشف الطبي هي AED ١٧.٨٢، يمكنكم الدفع عند الوصول"}, {"role": "assistant", "content": "تكلفة الكشف الطبي هي سبعة عشر درهماً واثنان وثمانون فلساً، يمكنكم الدفع عند الوصول"}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "سيصلك المندوب خلال ٦٥ دقائق، يرجى البقاء بالقرب من ٠١٥ ٨٣٦٥ ٦١٢٥."}, {"role": "assistant", "content": "سيصلك المندوب خلال خمسة وستون دقائق، يرجى البقاء بالقرب من صفر واحد خمسة ثمانية ثلاثة ستة خمسة ستة واحد اثنان خمسة."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تكلفة الكشف الطبي هي EGP 35.90، يمكنكم الدفع عند الوصول"}, {"role": "assistant", "content": "تكلفة الكشف الطبي هي خمسة وثلاثون جنيهاً وتسعون قرشاً، يمكنكم الدفع عند الوصول"}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "أرسل النموذج إلى maria@posta.it أو تفضل بزيارة https://sltnet.lk/faq."}, {"role": "assistant", "content": "أرسل النموذج إلى maria آت posta نقطة it أو تفضل بزيارة sltnet نقطة lk شرطة مائلة faq."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تكلفة الكشف الطبي هي ٥٨,١١٣.٥٠ دولار، يمكنكم الدفع عند الوصول"}, {"role": "assistant", "content": "تكلفة الكشف الطبي هي ثمانية وخمسون ألفاً ومائة وثلاثة عشر دولاراً وخمسون سنتاً، يمكنكم الدفع عند الوصول"}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا https://yahoo.com.ph/help."}, {"role": "assistant", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا yahoo نقطة com نقطة ph شرطة مائلة help."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "سيصلك المندوب خلال 21 دقائق، يرجى البقاء بالقرب من +966 55 011 8488."}, {"role": "assistant", "content": "سيصلك المندوب خلال واحد وعشرون دقائق، يرجى البقاء بالقرب من زائد تسعة ستة ستة خمسة خمسة صفر واحد واحد ثمانية أربعة ثمانية ثمانية."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "سيصلك المندوب خلال 999 دقائق، يرجى البقاء بالقرب من +966 59 142 7375."}, {"role": "assistant", "content": "سيصلك المندوب خلال تسعمائة وتسعة وتسعون دقائق، يرجى البقاء بالقرب من زائد تسعة ستة ستة خمسة تسعة واحد أربعة اثنان سبعة ثلاثة سبعة خمسة."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا www.yahoo.com/login."}, {"role": "assistant", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا w w w نقطة yahoo نقطة com شرطة مائلة login."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "للاستفسار عن حالة طلبك MH370، يرجى الاتصال بنا على 057 098 3593."}, {"role": "assistant", "content": "للاستفسار عن حالة طلبك MH ثلاثة سبعة صفر، يرجى الاتصال بنا على صفر خمسة سبعة صفر تسعة ثمانية ثلاثة خمسة تسعة ثلاثة."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "للاستفسار عن حالة طلبك MH370، يرجى الاتصال بنا على 056 655 7162."}, {"role": "assistant", "content": "للاستفسار عن حالة طلبك MH ثلاثة سبعة صفر، يرجى الاتصال بنا على صفر خمسة ستة ستة خمسة خمسة سبعة واحد ستة اثنان."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا https://www.gmail.co.id/support."}, {"role": "assistant", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا w w w نقطة gmail نقطة co نقطة id شرطة مائلة support."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "للاستفسار عن حالة طلبك WXY 1234، يرجى الاتصال بنا على 0583295138."}, {"role": "assistant", "content": "للاستفسار عن حالة طلبك WXY واحد اثنان ثلاثة أربعة، يرجى الاتصال بنا على صفر خمسة ثمانية ثلاثة اثنان تسعة خمسة واحد ثلاثة ثمانية."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تكلفة الكشف الطبي هي $332.43، يمكنكم الدفع عند الوصول"}, {"role": "assistant", "content": "تكلفة الكشف الطبي هي ثلاثمائة واثنان وثلاثون دولاراً وثلاثة وأربعون سنتاً، يمكنكم الدفع عند الوصول"}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا https://outlook.sa."}, {"role": "assistant", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا outlook نقطة sa."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تكلفة الكشف الطبي هي $47,902.50، يمكنكم الدفع عند الوصول"}, {"role": "assistant", "content": "تكلفة الكشف الطبي هي سبعة وأربعون ألفاً وتسعمائة واثنان دولاراً وخمسون سنتاً، يمكنكم الدفع عند الوصول"}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "سيصلك المندوب خلال ٦٥٣ دقائق، يرجى البقاء بالقرب من ٠٥٨ ٤٩٥ ٤٥٠٣."}, {"role": "assistant", "content": "سيصلك المندوب خلال ستمائة وثلاثة وخمسون دقائق، يرجى البقاء بالقرب من صفر خمسة ثمانية أربعة تسعة خمسة أربعة خمسة صفر ثلاثة."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا https://www.web.de/support."}, {"role": "assistant", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا w w w نقطة web نقطة de شرطة مائلة support."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "أرسل النموذج إلى siti@yahoo.com أو تفضل بزيارة www.sltnet.lk/login."}, {"role": "assistant", "content": "أرسل النموذج إلى siti آت yahoo نقطة com أو تفضل بزيارة w w w نقطة sltnet نقطة lk شرطة مائلة login."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تكلفة الكشف الطبي هي 895 درهم، يمكنكم الدفع عند الوصول"}, {"role": "assistant", "content": "تكلفة الكشف الطبي هي ثمانمائة وخمسة وتسعون درهماً، يمكنكم الدفع عند الوصول"}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تكلفة الكشف الطبي هي AED 118.50، يمكنكم الدفع عند الوصول"}, {"role": "assistant", "content": "تكلفة الكشف الطبي هي مائة وثمانية عشر درهماً وخمسون فلساً، يمكنكم الدفع عند الوصول"}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تكلفة الكشف الطبي هي 56 درهم، يمكنكم الدفع عند الوصول"}, {"role": "assistant", "content": "تكلفة الكشف الطبي هي ستة وخمسون درهماً، يمكنكم الدفع عند الوصول"}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تكلفة الكشف الطبي هي SAR ١٧.٩٠، يمكنكم الدفع عند الوصول"}, {"role": "assistant", "content": "تكلفة الكشف الطبي هي سبعة عشر ريالاً وتسعون هللةً، يمكنكم الدفع عند الوصول"}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "سيصلك المندوب خلال 979 دقائق، يرجى البقاء بالقرب من 016 2456 5740."}, {"role": "assistant", "content": "سيصلك المندوب خلال تسعمائة وتسعة وسبعون دقائق، يرجى البقاء بالقرب من صفر واحد ستة اثنان أربعة خمسة ستة خمسة سبعة أربعة صفر."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "أرسل النموذج إلى jan.devries@gmail.co.id أو تفضل بزيارة www.ziggo.nl/login."}, {"role": "assistant", "content": "أرسل النموذج إلى jan نقطة devries آت gmail نقطة co نقطة id أو تفضل بزيارة w w w نقطة ziggo نقطة nl شرطة مائلة login."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا www.gmail.com."}, {"role": "assistant", "content": "يمكنك تتبع مسار شحنتك مباشرة عبر موقعنا w w w نقطة gmail نقطة com."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "سيصلك المندوب خلال 7 دقائق، يرجى البقاء بالقرب من 015 1192 7792."}, {"role": "assistant", "content": "سيصلك المندوب خلال سبعة دقائق، يرجى البقاء بالقرب من صفر واحد خمسة واحد واحد تسعة اثنان سبعة سبعة تسعة اثنان."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "للاستفسار عن حالة طلبك REF-2024-0113، يرجى الاتصال بنا على 019 9022 8639."}, {"role": "assistant", "content": "للاستفسار عن حالة طلبك REF اثنان صفر اثنان أربعة صفر واحد واحد ثلاثة، يرجى الاتصال بنا على صفر واحد تسعة تسعة صفر اثنان اثنان ثمانية ستة ثلاثة تسعة."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "للاستفسار عن حالة طلبك AK6120، يرجى الاتصال بنا على 0578811618."}, {"role": "assistant", "content": "للاستفسار عن حالة طلبك AK ستة واحد اثنان صفر، يرجى الاتصال بنا على صفر خمسة سبعة ثمانية ثمانية واحد واحد ستة واحد ثمانية."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "للاستفسار عن حالة طلبك PO 55219، يرجى الاتصال بنا على +966 51 362 8960."}, {"role": "assistant", "content": "للاستفسار عن حالة طلبك PO خمسة خمسة اثنان واحد تسعة، يرجى الاتصال بنا على زائد تسعة ستة ستة خمسة واحد ثلاثة ستة اثنان ثمانية تسعة ستة صفر."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "أرسل النموذج إلى sofia@sapo.pt أو تفضل بزيارة https://mail.fr/faq."}, {"role": "assistant", "content": "أرسل النموذج إلى sofia آت sapo نقطة pt أو تفضل بزيارة mail نقطة fr شرطة مائلة faq."}], "lang": "ar", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez régler la somme de 975,85 € avant le 06/05/2027 pour éviter toute pénalité."}, {"role": "assistant", "content": "Veuillez régler la somme de neuf cent soixante-quinze euros et quatre-vingt-cinq centimes avant le six mai deux mille vingt-sept pour éviter toute pénalité."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il reste 47 places disponibles pour le séminaire du 19 janvier 2031."}, {"role": "assistant", "content": "Il reste quarante-sept places disponibles pour le séminaire du dix-neuf janvier deux mille trente et un."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez confirmer votre identité en saisissant le code 99793 reçu par SMS."}, {"role": "assistant", "content": "Veuillez confirmer votre identité en saisissant le code neuf neuf sept neuf trois reçu par SMS."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Votre code est le 418547, valable 72 minutes."}, {"role": "assistant", "content": "Votre code est le quatre un huit cinq quatre sept, valable soixante-douze minutes."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il reste 78 places disponibles pour le séminaire du 27 juin."}, {"role": "assistant", "content": "Il reste soixante-dix-huit places disponibles pour le séminaire du vingt-sept juin."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez confirmer votre identité en saisissant le code 4787 reçu par SMS."}, {"role": "assistant", "content": "Veuillez confirmer votre identité en saisissant le code quatre sept huit sept reçu par SMS."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez régler la somme de 393,90 € avant le 11 janvier pour éviter toute pénalité."}, {"role": "assistant", "content": "Veuillez régler la somme de trois cent quatre-vingt-treize euros et quatre-vingt-dix centimes avant le onze janvier pour éviter toute pénalité."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il reste 43 places disponibles pour le séminaire du 21 mars."}, {"role": "assistant", "content": "Il reste quarante-trois places disponibles pour le séminaire du vingt et un mars."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 16,13 sur 47."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de seize virgule un trois sur quarante-sept."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 75,44 sur 183."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de soixante-quinze virgule quatre quatre sur cent quatre-vingt-trois."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Votre code est le 0982, valable 21 minutes."}, {"role": "assistant", "content": "Votre code est le zéro neuf huit deux, valable vingt et un minutes."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Votre code est le 076881, valable 672 minutes."}, {"role": "assistant", "content": "Votre code est le zéro sept six huit huit un, valable six cent soixante-douze minutes."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Votre code est le 93209, valable 66 minutes."}, {"role": "assistant", "content": "Votre code est le neuf trois deux zéro neuf, valable soixante-six minutes."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 53,19 sur 9."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de cinquante-trois virgule un neuf sur neuf."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez confirmer votre identité en saisissant le code 358638 reçu par SMS."}, {"role": "assistant", "content": "Veuillez confirmer votre identité en saisissant le code trois cinq huit six trois huit reçu par SMS."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il reste 896 places disponibles pour le séminaire du 25 février."}, {"role": "assistant", "content": "Il reste huit cent quatre-vingt-seize places disponibles pour le séminaire du vingt-cinq février."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il reste 62 places disponibles pour le séminaire du 05/09/2008."}, {"role": "assistant", "content": "Il reste soixante-deux places disponibles pour le séminaire du cinq septembre deux mille huit."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 56,81 sur 52."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de cinquante-six virgule huit un sur cinquante-deux."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez régler la somme de 850,56 € avant le 23 août pour éviter toute pénalité."}, {"role": "assistant", "content": "Veuillez régler la somme de huit cent cinquante euros et cinquante-six centimes avant le vingt-trois août pour éviter toute pénalité."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez régler la somme de EUR 31 519,50 avant le 28/12/2028 pour éviter toute pénalité."}, {"role": "assistant", "content": "Veuillez régler la somme de trente et un mille cinq cent dix-neuf euros et cinquante centimes avant le vingt-huit décembre deux mille vingt-huit pour éviter toute pénalité."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez régler la somme de 217€ avant le 26 janvier 2029 pour éviter toute pénalité."}, {"role": "assistant", "content": "Veuillez régler la somme de deux cent dix-sept euros avant le vingt-six janvier deux mille vingt-neuf pour éviter toute pénalité."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez régler la somme de EUR 50 529 avant le 7/5/1997 pour éviter toute pénalité."}, {"role": "assistant", "content": "Veuillez régler la somme de cinquante mille cinq cent vingt-neuf euros avant le sept mai mille neuf cent quatre-vingt-dix-sept pour éviter toute pénalité."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Votre code est le 67590, valable 8 minutes."}, {"role": "assistant", "content": "Votre code est le six sept cinq neuf zéro, valable huit minutes."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Votre code est le 28581, valable 952 minutes."}, {"role": "assistant", "content": "Votre code est le deux huit cinq huit un, valable neuf cent cinquante-deux minutes."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Votre code est le 32146, valable 9 minutes."}, {"role": "assistant", "content": "Votre code est le trois deux un quatre six, valable neuf minutes."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez confirmer votre identité en saisissant le code 09612 reçu par SMS."}, {"role": "assistant", "content": "Veuillez confirmer votre identité en saisissant le code zéro neuf six un deux reçu par SMS."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez régler la somme de 46,87 € avant le 16/3/2025 pour éviter toute pénalité."}, {"role": "assistant", "content": "Veuillez régler la somme de quarante-six euros et quatre-vingt-sept centimes avant le seize mars deux mille vingt-cinq pour éviter toute pénalité."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Votre code est le 6301, valable 28 minutes."}, {"role": "assistant", "content": "Votre code est le six trois zéro un, valable vingt-huit minutes."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Votre code est le 067955, valable 43 minutes."}, {"role": "assistant", "content": "Votre code est le zéro six sept neuf cinq cinq, valable quarante-trois minutes."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il reste 58 places disponibles pour le séminaire du 8 janvier."}, {"role": "assistant", "content": "Il reste cinquante-huit places disponibles pour le séminaire du huit janvier."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez confirmer votre identité en saisissant le code 8096 reçu par SMS."}, {"role": "assistant", "content": "Veuillez confirmer votre identité en saisissant le code huit zéro neuf six reçu par SMS."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 62,94 sur 9."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de soixante-deux virgule neuf quatre sur neuf."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il reste 97 places disponibles pour le séminaire du 2/3/2003."}, {"role": "assistant", "content": "Il reste quatre-vingt-dix-sept places disponibles pour le séminaire du deux mars deux mille trois."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il reste 159 places disponibles pour le séminaire du 12 décembre 1999."}, {"role": "assistant", "content": "Il reste cent cinquante-neuf places disponibles pour le séminaire du douze décembre mille neuf cent quatre-vingt-dix-neuf."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 45,08 sur 629."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de quarante-cinq virgule zéro huit sur six cent vingt-neuf."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Votre code est le 455555, valable 83 minutes."}, {"role": "assistant", "content": "Votre code est le quatre cinq cinq cinq cinq cinq, valable quatre-vingt-trois minutes."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il reste 835 places disponibles pour le séminaire du 7 janvier 2000."}, {"role": "assistant", "content": "Il reste huit cent trente-cinq places disponibles pour le séminaire du sept janvier deux mille."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il reste 296 places disponibles pour le séminaire du 24 avril."}, {"role": "assistant", "content": "Il reste deux cent quatre-vingt-seize places disponibles pour le séminaire du vingt-quatre avril."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez régler la somme de 49€ avant le 24 novembre pour éviter toute pénalité."}, {"role": "assistant", "content": "Veuillez régler la somme de quarante-neuf euros avant le vingt-quatre novembre pour éviter toute pénalité."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez régler la somme de 649 € avant le 4/7/2023 pour éviter toute pénalité."}, {"role": "assistant", "content": "Veuillez régler la somme de six cent quarante-neuf euros avant le quatre juillet deux mille vingt-trois pour éviter toute pénalité."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 17,1 sur 4."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de dix-sept virgule un sur quatre."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 63,2 sur 9."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de soixante-trois virgule deux sur neuf."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez régler la somme de EUR 75 avant le 11/11/2018 pour éviter toute pénalité."}, {"role": "assistant", "content": "Veuillez régler la somme de soixante-quinze euros avant le onze novembre deux mille dix-huit pour éviter toute pénalité."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Votre code est le 66752, valable 644 minutes."}, {"role": "assistant", "content": "Votre code est le six six sept cinq deux, valable six cent quarante-quatre minutes."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez régler la somme de EUR 66,50 avant le 23/08/1999 pour éviter toute pénalité."}, {"role": "assistant", "content": "Veuillez régler la somme de soixante-six euros et cinquante centimes avant le vingt-trois août mille neuf cent quatre-vingt-dix-neuf pour éviter toute pénalité."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez régler la somme de 38 957,90€ avant le 26/10/2017 pour éviter toute pénalité."}, {"role": "assistant", "content": "Veuillez régler la somme de trente-huit mille neuf cent cinquante-sept euros et quatre-vingt-dix centimes avant le vingt-six octobre deux mille dix-sept pour éviter toute pénalité."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 69,86 sur 5."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de soixante-neuf virgule huit six sur cinq."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 34,11 sur 82."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de trente-quatre virgule un un sur quatre-vingt-deux."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il reste 5 places disponibles pour le séminaire du 8 janvier."}, {"role": "assistant", "content": "Il reste cinq places disponibles pour le séminaire du huit janvier."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez régler la somme de 90,55€ avant le 5/12/2001 pour éviter toute pénalité."}, {"role": "assistant", "content": "Veuillez régler la somme de quatre-vingt-dix euros et cinquante-cinq centimes avant le cinq décembre deux mille un pour éviter toute pénalité."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Votre code est le 7484, valable 288 minutes."}, {"role": "assistant", "content": "Votre code est le sept quatre huit quatre, valable deux cent quatre-vingt-huit minutes."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 64,3 sur 88."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de soixante-quatre virgule trois sur quatre-vingt-huit."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 6,8 sur 80."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de six virgule huit sur quatre-vingts."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 86,2 sur 847."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de quatre-vingt-six virgule deux sur huit cent quarante-sept."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 95,9 sur 83."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de quatre-vingt-quinze virgule neuf sur quatre-vingt-trois."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez régler la somme de 6 829€ avant le 10/04/2003 pour éviter toute pénalité."}, {"role": "assistant", "content": "Veuillez régler la somme de six mille huit cent vingt-neuf euros avant le dix avril deux mille trois pour éviter toute pénalité."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez confirmer votre identité en saisissant le code 28099 reçu par SMS."}, {"role": "assistant", "content": "Veuillez confirmer votre identité en saisissant le code deux huit zéro neuf neuf reçu par SMS."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Votre code est le 3594, valable 3 minutes."}, {"role": "assistant", "content": "Votre code est le trois cinq neuf quatre, valable trois minutes."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez confirmer votre identité en saisissant le code 15803 reçu par SMS."}, {"role": "assistant", "content": "Veuillez confirmer votre identité en saisissant le code un cinq huit zéro trois reçu par SMS."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez régler la somme de EUR 440,94 avant le 27/05/1997 pour éviter toute pénalité."}, {"role": "assistant", "content": "Veuillez régler la somme de quatre cent quarante euros et quatre-vingt-quatorze centimes avant le vingt-sept mai mille neuf cent quatre-vingt-dix-sept pour éviter toute pénalité."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez confirmer votre identité en saisissant le code 5446 reçu par SMS."}, {"role": "assistant", "content": "Veuillez confirmer votre identité en saisissant le code cinq quatre quatre six reçu par SMS."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 48,63 sur 2."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de quarante-huit virgule six trois sur deux."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Votre code est le 341678, valable 64 minutes."}, {"role": "assistant", "content": "Votre code est le trois quatre un six sept huit, valable soixante-quatre minutes."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Votre code est le 316646, valable 6 minutes."}, {"role": "assistant", "content": "Votre code est le trois un six six quatre six, valable six minutes."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il reste 39 places disponibles pour le séminaire du 12 mars."}, {"role": "assistant", "content": "Il reste trente-neuf places disponibles pour le séminaire du douze mars."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 17,7 sur 2."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de dix-sept virgule sept sur deux."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 27,1 sur 4."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de vingt-sept virgule un sur quatre."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez confirmer votre identité en saisissant le code 345475 reçu par SMS."}, {"role": "assistant", "content": "Veuillez confirmer votre identité en saisissant le code trois quatre cinq quatre sept cinq reçu par SMS."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il reste 4 places disponibles pour le séminaire du 28 avril 2031."}, {"role": "assistant", "content": "Il reste quatre places disponibles pour le séminaire du vingt-huit avril deux mille trente et un."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez confirmer votre identité en saisissant le code 258017 reçu par SMS."}, {"role": "assistant", "content": "Veuillez confirmer votre identité en saisissant le code deux cinq huit zéro un sept reçu par SMS."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez confirmer votre identité en saisissant le code 81574 reçu par SMS."}, {"role": "assistant", "content": "Veuillez confirmer votre identité en saisissant le code huit un cinq sept quatre reçu par SMS."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il reste 460 places disponibles pour le séminaire du 23 mai 2027."}, {"role": "assistant", "content": "Il reste quatre cent soixante places disponibles pour le séminaire du vingt-trois mai deux mille vingt-sept."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 21,97 sur 9."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de vingt et un virgule neuf sept sur neuf."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Votre code est le 051326, valable 426 minutes."}, {"role": "assistant", "content": "Votre code est le zéro cinq un trois deux six, valable quatre cent vingt-six minutes."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez régler la somme de 11 € avant le 4 mai pour éviter toute pénalité."}, {"role": "assistant", "content": "Veuillez régler la somme de onze euros avant le quatre mai pour éviter toute pénalité."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 43,10 sur 2."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de quarante-trois virgule un zéro sur deux."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez régler la somme de 62,50€ avant le 24/02/2032 pour éviter toute pénalité."}, {"role": "assistant", "content": "Veuillez régler la somme de soixante-deux euros et cinquante centimes avant le vingt-quatre février deux mille trente-deux pour éviter toute pénalité."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez confirmer votre identité en saisissant le code 0460 reçu par SMS."}, {"role": "assistant", "content": "Veuillez confirmer votre identité en saisissant le code zéro quatre six zéro reçu par SMS."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Votre code est le 02095, valable 42 minutes."}, {"role": "assistant", "content": "Votre code est le zéro deux zéro neuf cinq, valable quarante-deux minutes."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez confirmer votre identité en saisissant le code 612513 reçu par SMS."}, {"role": "assistant", "content": "Veuillez confirmer votre identité en saisissant le code six un deux cinq un trois reçu par SMS."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez confirmer votre identité en saisissant le code 995760 reçu par SMS."}, {"role": "assistant", "content": "Veuillez confirmer votre identité en saisissant le code neuf neuf cinq sept six zéro reçu par SMS."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Votre code est le 13348, valable 8 minutes."}, {"role": "assistant", "content": "Votre code est le un trois trois quatre huit, valable huit minutes."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 45,6 sur 3."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de quarante-cinq virgule six sur trois."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 85,92 sur 19."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de quatre-vingt-cinq virgule neuf deux sur dix-neuf."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 24,43 sur 9."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de vingt-quatre virgule quatre trois sur neuf."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Votre code est le 361025, valable 45 minutes."}, {"role": "assistant", "content": "Votre code est le trois six un zéro deux cinq, valable quarante-cinq minutes."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 79,8 sur 573."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de soixante-dix-neuf virgule huit sur cinq cent soixante-treize."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez confirmer votre identité en saisissant le code 48459 reçu par SMS."}, {"role": "assistant", "content": "Veuillez confirmer votre identité en saisissant le code quatre huit quatre cinq neuf reçu par SMS."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Votre code est le 0706, valable 7 minutes."}, {"role": "assistant", "content": "Votre code est le zéro sept zéro six, valable sept minutes."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez régler la somme de EUR 99 avant le 15 juin pour éviter toute pénalité."}, {"role": "assistant", "content": "Veuillez régler la somme de quatre-vingt-dix-neuf euros avant le quinze juin pour éviter toute pénalité."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez confirmer votre identité en saisissant le code 2391 reçu par SMS."}, {"role": "assistant", "content": "Veuillez confirmer votre identité en saisissant le code deux trois neuf un reçu par SMS."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez régler la somme de 13 € avant le 8 novembre pour éviter toute pénalité."}, {"role": "assistant", "content": "Veuillez régler la somme de treize euros avant le huit novembre pour éviter toute pénalité."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 64,69 sur 24."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de soixante-quatre virgule six neuf sur vingt-quatre."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 80,1 sur 436."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de quatre-vingts virgule un sur quatre cent trente-six."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 57,61 sur 8."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de cinquante-sept virgule six un sur huit."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il reste 3 places disponibles pour le séminaire du 06/09/2020."}, {"role": "assistant", "content": "Il reste trois places disponibles pour le séminaire du six septembre deux mille vingt."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il reste 294 places disponibles pour le séminaire du 13 août 2004."}, {"role": "assistant", "content": "Il reste deux cent quatre-vingt-quatorze places disponibles pour le séminaire du treize août deux mille quatre."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez confirmer votre identité en saisissant le code 659281 reçu par SMS."}, {"role": "assistant", "content": "Veuillez confirmer votre identité en saisissant le code six cinq neuf deux huit un reçu par SMS."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez confirmer votre identité en saisissant le code 07676 reçu par SMS."}, {"role": "assistant", "content": "Veuillez confirmer votre identité en saisissant le code zéro sept six sept six reçu par SMS."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez régler la somme de 1 605,50€ avant le 17/10/2010 pour éviter toute pénalité."}, {"role": "assistant", "content": "Veuillez régler la somme de mille six cent cinq euros et cinquante centimes avant le dix-sept octobre deux mille dix pour éviter toute pénalité."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 71,87 sur 54."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de soixante et onze virgule huit sept sur cinquante-quatre."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Votre code est le 85667, valable 14 minutes."}, {"role": "assistant", "content": "Votre code est le huit cinq six six sept, valable quatorze minutes."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 76,29 sur 7."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de soixante-seize virgule deux neuf sur sept."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez confirmer votre identité en saisissant le code 87515 reçu par SMS."}, {"role": "assistant", "content": "Veuillez confirmer votre identité en saisissant le code huit sept cinq un cinq reçu par SMS."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez confirmer votre identité en saisissant le code 273620 reçu par SMS."}, {"role": "assistant", "content": "Veuillez confirmer votre identité en saisissant le code deux sept trois six deux zéro reçu par SMS."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 81,71 sur 33."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de quatre-vingt-un virgule sept un sur trente-trois."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Votre code est le 90902, valable 53 minutes."}, {"role": "assistant", "content": "Votre code est le neuf zéro neuf zéro deux, valable cinquante-trois minutes."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 81,54 sur 499."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de quatre-vingt-un virgule cinq quatre sur quatre cent quatre-vingt-dix-neuf."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez régler la somme de 95,90€ avant le 8 décembre 2024 pour éviter toute pénalité."}, {"role": "assistant", "content": "Veuillez régler la somme de quatre-vingt-quinze euros et quatre-vingt-dix centimes avant le huit décembre deux mille vingt-quatre pour éviter toute pénalité."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez régler la somme de EUR 36,50 avant le 15/1/2023 pour éviter toute pénalité."}, {"role": "assistant", "content": "Veuillez régler la somme de trente-six euros et cinquante centimes avant le quinze janvier deux mille vingt-trois pour éviter toute pénalité."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Votre code est le 49752, valable 184 minutes."}, {"role": "assistant", "content": "Votre code est le quatre neuf sept cinq deux, valable cent quatre-vingt-quatre minutes."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il reste 5 places disponibles pour le séminaire du 22/8/2006."}, {"role": "assistant", "content": "Il reste cinq places disponibles pour le séminaire du vingt-deux août deux mille six."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez régler la somme de 240 € avant le 24/01/2007 pour éviter toute pénalité."}, {"role": "assistant", "content": "Veuillez régler la somme de deux cent quarante euros avant le vingt-quatre janvier deux mille sept pour éviter toute pénalité."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il reste 96 places disponibles pour le séminaire du 1/2/2031."}, {"role": "assistant", "content": "Il reste quatre-vingt-seize places disponibles pour le séminaire du premier février deux mille trente et un."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Votre code est le 7810, valable 5 minutes."}, {"role": "assistant", "content": "Votre code est le sept huit un zéro, valable cinq minutes."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez confirmer votre identité en saisissant le code 98496 reçu par SMS."}, {"role": "assistant", "content": "Veuillez confirmer votre identité en saisissant le code neuf huit quatre neuf six reçu par SMS."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 91,4 sur 5."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de quatre-vingt-onze virgule quatre sur cinq."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Votre code est le 8508, valable 74 minutes."}, {"role": "assistant", "content": "Votre code est le huit cinq zéro huit, valable soixante-quatorze minutes."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez confirmer votre identité en saisissant le code 866216 reçu par SMS."}, {"role": "assistant", "content": "Veuillez confirmer votre identité en saisissant le code huit six six deux un six reçu par SMS."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 56,9 sur 49."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de cinquante-six virgule neuf sur quarante-neuf."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il reste 18 places disponibles pour le séminaire du 24 avril 2023."}, {"role": "assistant", "content": "Il reste dix-huit places disponibles pour le séminaire du vingt-quatre avril deux mille vingt-trois."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez confirmer votre identité en saisissant le code 833651 reçu par SMS."}, {"role": "assistant", "content": "Veuillez confirmer votre identité en saisissant le code huit trois trois six cinq un reçu par SMS."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez régler la somme de 13 467,70€ avant le 26/02/2000 pour éviter toute pénalité."}, {"role": "assistant", "content": "Veuillez régler la somme de treize mille quatre cent soixante-sept euros et soixante-dix centimes avant le vingt-six février deux mille pour éviter toute pénalité."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 24,5 sur 4."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de vingt-quatre virgule cinq sur quatre."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Votre code est le 994061, valable 627 minutes."}, {"role": "assistant", "content": "Votre code est le neuf neuf quatre zéro six un, valable six cent vingt-sept minutes."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez régler la somme de 84,30€ avant le 27/12/2006 pour éviter toute pénalité."}, {"role": "assistant", "content": "Veuillez régler la somme de quatre-vingt-quatre euros et trente centimes avant le vingt-sept décembre deux mille six pour éviter toute pénalité."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il reste 291 places disponibles pour le séminaire du 25/1/1999."}, {"role": "assistant", "content": "Il reste deux cent quatre-vingt-onze places disponibles pour le séminaire du vingt-cinq janvier mille neuf cent quatre-vingt-dix-neuf."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez régler la somme de 613,65€ avant le 14/11/2004 pour éviter toute pénalité."}, {"role": "assistant", "content": "Veuillez régler la somme de six cent treize euros et soixante-cinq centimes avant le quatorze novembre deux mille quatre pour éviter toute pénalité."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez régler la somme de 27,90€ avant le 28/01/2030 pour éviter toute pénalité."}, {"role": "assistant", "content": "Veuillez régler la somme de vingt-sept euros et quatre-vingt-dix centimes avant le vingt-huit janvier deux mille trente pour éviter toute pénalité."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 98,24 sur 4."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de quatre-vingt-dix-huit virgule deux quatre sur quatre."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez régler la somme de 265,75€ avant le 13/2/2012 pour éviter toute pénalité."}, {"role": "assistant", "content": "Veuillez régler la somme de deux cent soixante-cinq euros et soixante-quinze centimes avant le treize février deux mille douze pour éviter toute pénalité."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il reste 114 places disponibles pour le séminaire du 14/2/1999."}, {"role": "assistant", "content": "Il reste cent quatorze places disponibles pour le séminaire du quatorze février mille neuf cent quatre-vingt-dix-neuf."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Veuillez confirmer votre identité en saisissant le code 04073 reçu par SMS."}, {"role": "assistant", "content": "Veuillez confirmer votre identité en saisissant le code zéro quatre zéro sept trois reçu par SMS."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 14,21 sur 398."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de quatorze virgule deux un sur trois cent quatre-vingt-dix-huit."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La moyenne requise pour valider le module est de 89,1 sur 8."}, {"role": "assistant", "content": "La moyenne requise pour valider le module est de quatre-vingt-neuf virgule un sur huit."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Votre code est le 87259, valable 5 minutes."}, {"role": "assistant", "content": "Votre code est le huit sept deux cinq neuf, valable cinq minutes."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il reste 88 places disponibles pour le séminaire du 3 janvier."}, {"role": "assistant", "content": "Il reste quatre-vingt-huit places disponibles pour le séminaire du trois janvier."}], "lang": "fr", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de 65,50 € en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de sesenta y cinco euros con cincuenta céntimos en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 05/12/2007 del año 1992."}, {"role": "assistant", "content": "La cuenta fue abierta el cinco de diciembre de dos mil siete del año mil novecientos noventa y dos."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para más información sobre su envío REF-2024-0113, visite nuestro portal en www.ziggo.nl."}, {"role": "assistant", "content": "Para más información sobre su envío REF dos cero dos cuatro cero uno uno tres, visite nuestro portal en w w w punto ziggo punto nl."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para más información sobre su envío X-12340567, visite nuestro portal en https://correo.es."}, {"role": "assistant", "content": "Para más información sobre su envío X uno dos tres cuatro cero cinco seis siete, visite nuestro portal en correo punto es."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para gestionar su devolución, acceda a https://ziggo.nl/help con su número de pedido QZ8501."}, {"role": "assistant", "content": "Para gestionar su devolución, acceda a ziggo punto nl barra help con su número de pedido QZ ocho cinco cero uno."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para gestionar su devolución, acceda a www.yahoo.com.ph/login con su número de pedido INV0472."}, {"role": "assistant", "content": "Para gestionar su devolución, acceda a w w w punto yahoo punto com punto ph barra login con su número de pedido INV cero cuatro siete dos."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de 733€ en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de setecientos treinta y tres euros en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hemos detectado un error en su dirección; por favor llámenos al 728 353 158."}, {"role": "assistant", "content": "Hemos detectado un error en su dirección; por favor llámenos al siete dos ocho tres cinco tres uno cinco ocho."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 1/9/2014 del año 1957."}, {"role": "assistant", "content": "La cuenta fue abierta el uno de septiembre de dos mil catorce del año mil novecientos cincuenta y siete."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para gestionar su devolución, acceda a https://gmail.co.id/help con su número de pedido A12."}, {"role": "assistant", "content": "Para gestionar su devolución, acceda a gmail punto co punto id barra help con su número de pedido A uno dos."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para gestionar su devolución, acceda a https://outlook.sa/faq con su número de pedido AK6120."}, {"role": "assistant", "content": "Para gestionar su devolución, acceda a outlook punto sa barra faq con su número de pedido AK seis uno dos cero."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hemos detectado un error en su dirección; por favor llámenos al 629 060 035."}, {"role": "assistant", "content": "Hemos detectado un error en su dirección; por favor llámenos al seis dos nueve cero seis cero cero tres cinco."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de 9.122,50€ en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de nueve mil ciento veintidós euros con cincuenta céntimos en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para más información sobre su envío AK6120, visite nuestro portal en https://web.de/help."}, {"role": "assistant", "content": "Para más información sobre su envío AK seis uno dos cero, visite nuestro portal en web punto de barra help."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 15/9/2014 del año 1980."}, {"role": "assistant", "content": "La cuenta fue abierta el quince de septiembre de dos mil catorce del año mil novecientos ochenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para gestionar su devolución, acceda a www.yahoo.com.ph/login con su número de pedido A12."}, {"role": "assistant", "content": "Para gestionar su devolución, acceda a w w w punto yahoo punto com punto ph barra login con su número de pedido A uno dos."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de EUR 754,90 en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de setecientos cincuenta y cuatro euros con noventa céntimos en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para gestionar su devolución, acceda a https://sltnet.lk con su número de pedido ABC1234."}, {"role": "assistant", "content": "Para gestionar su devolución, acceda a sltnet punto lk con su número de pedido ABC uno dos tres cuatro."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para gestionar su devolución, acceda a https://www.sapo.pt/support con su número de pedido REF-2024-0113."}, {"role": "assistant", "content": "Para gestionar su devolución, acceda a w w w punto sapo punto pt barra support con su número de pedido REF dos cero dos cuatro cero uno uno tres."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 6 de marzo del año 2010."}, {"role": "assistant", "content": "La cuenta fue abierta el seis de marzo del año dos mil diez."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 16/03/2010 del año 1960."}, {"role": "assistant", "content": "La cuenta fue abierta el dieciséis de marzo de dos mil diez del año mil novecientos sesenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de EUR 74,19 en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de setenta y cuatro euros con diecinueve céntimos en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para más información sobre su envío TK-88213, visite nuestro portal en https://www.company.co/support."}, {"role": "assistant", "content": "Para más información sobre su envío TK ocho ocho dos uno tres, visite nuestro portal en w w w punto company punto co barra support."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para gestionar su devolución, acceda a https://sltnet.lk/faq con su número de pedido X-12340567."}, {"role": "assistant", "content": "Para gestionar su devolución, acceda a sltnet punto lk barra faq con su número de pedido X uno dos tres cuatro cero cinco seis siete."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hemos detectado un error en su dirección; por favor llámenos al 966 664 941."}, {"role": "assistant", "content": "Hemos detectado un error en su dirección; por favor llámenos al nueve seis seis seis seis cuatro nueve cuatro uno."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de EUR 5.540,90 en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de cinco mil quinientos cuarenta euros con noventa céntimos en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para más información sobre su envío PO 55219, visite nuestro portal en https://yahoo.com."}, {"role": "assistant", "content": "Para más información sobre su envío PO cinco cinco dos uno nueve, visite nuestro portal en yahoo punto com."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hemos detectado un error en su dirección; por favor llámenos al 906 185 323."}, {"role": "assistant", "content": "Hemos detectado un error en su dirección; por favor llámenos al nueve cero seis uno ocho cinco tres dos tres."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hemos detectado un error en su dirección; por favor llámenos al 910 804 109."}, {"role": "assistant", "content": "Hemos detectado un error en su dirección; por favor llámenos al nueve uno cero ocho cero cuatro uno cero nueve."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hemos detectado un error en su dirección; por favor llámenos al 632 900 162."}, {"role": "assistant", "content": "Hemos detectado un error en su dirección; por favor llámenos al seis tres dos nueve cero cero uno seis dos."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de 22.930,21 € en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de veintidós mil novecientos treinta euros con veintiuno céntimos en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para más información sobre su envío ABC1234, visite nuestro portal en www.posta.it/login."}, {"role": "assistant", "content": "Para más información sobre su envío ABC uno dos tres cuatro, visite nuestro portal en w w w punto posta punto it barra login."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 10/10/2020 del año 1958."}, {"role": "assistant", "content": "La cuenta fue abierta el diez de octubre de dos mil veinte del año mil novecientos cincuenta y ocho."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hemos detectado un error en su dirección; por favor llámenos al 641 760 218."}, {"role": "assistant", "content": "Hemos detectado un error en su dirección; por favor llámenos al seis cuatro uno siete seis cero dos uno ocho."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para más información sobre su envío MH370, visite nuestro portal en https://www.outlook.sa/support."}, {"role": "assistant", "content": "Para más información sobre su envío MH tres siete cero, visite nuestro portal en w w w punto outlook punto sa barra support."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 14 de enero del año 1995."}, {"role": "assistant", "content": "La cuenta fue abierta el catorce de enero del año mil novecientos noventa y cinco."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hemos detectado un error en su dirección; por favor llámenos al +34 696 418 200."}, {"role": "assistant", "content": "Hemos detectado un error en su dirección; por favor llámenos al más tres cuatro seis nueve seis cuatro uno ocho dos cero cero."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para gestionar su devolución, acceda a https://company.co con su número de pedido B2-14."}, {"role": "assistant", "content": "Para gestionar su devolución, acceda a company punto co con su número de pedido B dos uno cuatro."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para más información sobre su envío A12, visite nuestro portal en https://web.de/faq."}, {"role": "assistant", "content": "Para más información sobre su envío A uno dos, visite nuestro portal en web punto de barra faq."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de 478€ en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de cuatrocientos setenta y ocho euros en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de 978,66€ en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de novecientos setenta y ocho euros con sesenta y seis céntimos en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 11 de noviembre del año 1982."}, {"role": "assistant", "content": "La cuenta fue abierta el once de noviembre del año mil novecientos ochenta y dos."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de EUR 31.980 en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de treinta y uno mil novecientos ochenta euros en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hemos detectado un error en su dirección; por favor llámenos al 988 464 657."}, {"role": "assistant", "content": "Hemos detectado un error en su dirección; por favor llámenos al nueve ocho ocho cuatro seis cuatro seis cinco siete."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 7/12/2027 del año 2024."}, {"role": "assistant", "content": "La cuenta fue abierta el siete de diciembre de dos mil veintisiete del año dos mil veinticuatro."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para gestionar su devolución, acceda a https://gmail.co.id con su número de pedido REF-2024-0113."}, {"role": "assistant", "content": "Para gestionar su devolución, acceda a gmail punto co punto id con su número de pedido REF dos cero dos cuatro cero uno uno tres."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de 8.350€ en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de ocho mil trescientos cincuenta euros en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hemos detectado un error en su dirección; por favor llámenos al 733 024 260."}, {"role": "assistant", "content": "Hemos detectado un error en su dirección; por favor llámenos al siete tres tres cero dos cuatro dos seis cero."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de 97,50 € en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de noventa y siete euros con cincuenta céntimos en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 14/10/2005 del año 2018."}, {"role": "assistant", "content": "La cuenta fue abierta el catorce de octubre de dos mil cinco del año dos mil dieciocho."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 25/4/2008 del año 2028."}, {"role": "assistant", "content": "La cuenta fue abierta el veinticinco de abril de dos mil ocho del año dos mil veintiocho."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de 896€ en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de ochocientos noventa y seis euros en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 2 de septiembre de 2027 del año 1962."}, {"role": "assistant", "content": "La cuenta fue abierta el dos de septiembre de dos mil veintisiete del año mil novecientos sesenta y dos."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 16/05/2010 del año 2029."}, {"role": "assistant", "content": "La cuenta fue abierta el dieciséis de mayo de dos mil diez del año dos mil veintinueve."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de EUR 16 en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de dieciséis euros en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hemos detectado un error en su dirección; por favor llámenos al 909 548 065."}, {"role": "assistant", "content": "Hemos detectado un error en su dirección; por favor llámenos al nueve cero nueve cinco cuatro ocho cero seis cinco."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para gestionar su devolución, acceda a www.yahoo.com/login con su número de pedido A12."}, {"role": "assistant", "content": "Para gestionar su devolución, acceda a w w w punto yahoo punto com barra login con su número de pedido A uno dos."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para más información sobre su envío A12, visite nuestro portal en https://www.yahoo.com/support."}, {"role": "assistant", "content": "Para más información sobre su envío A uno dos, visite nuestro portal en w w w punto yahoo punto com barra support."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para más información sobre su envío TK-88213, visite nuestro portal en www.gmail.co.id/login."}, {"role": "assistant", "content": "Para más información sobre su envío TK ocho ocho dos uno tres, visite nuestro portal en w w w punto gmail punto co punto id barra login."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de 95.794 € en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de noventa y cinco mil setecientos noventa y cuatro euros en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 11/5/2024 del año 2008."}, {"role": "assistant", "content": "La cuenta fue abierta el once de mayo de dos mil veinticuatro del año dos mil ocho."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 24/02/2029 del año 1979."}, {"role": "assistant", "content": "La cuenta fue abierta el veinticuatro de febrero de dos mil veintinueve del año mil novecientos setenta y nueve."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de EUR 116,50 en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de ciento dieciséis euros con cincuenta céntimos en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de 9.785 € en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de nueve mil setecientos ochenta y cinco euros en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 21 de febrero de 2010 del año 1999."}, {"role": "assistant", "content": "La cuenta fue abierta el veintiuno de febrero de dos mil diez del año mil novecientos noventa y nueve."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 7 de junio del año 1986."}, {"role": "assistant", "content": "La cuenta fue abierta el siete de junio del año mil novecientos ochenta y seis."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para gestionar su devolución, acceda a www.company.co con su número de pedido B2-14."}, {"role": "assistant", "content": "Para gestionar su devolución, acceda a w w w punto company punto co con su número de pedido B dos uno cuatro."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de 66,68€ en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de sesenta y seis euros con sesenta y ocho céntimos en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 14/10/1995 del año 2001."}, {"role": "assistant", "content": "La cuenta fue abierta el catorce de octubre de mil novecientos noventa y cinco del año dos mil uno."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hemos detectado un error en su dirección; por favor llámenos al 719 012 292."}, {"role": "assistant", "content": "Hemos detectado un error en su dirección; por favor llámenos al siete uno nueve cero uno dos dos nueve dos."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 27 de abril del año 1975."}, {"role": "assistant", "content": "La cuenta fue abierta el veintisiete de abril del año mil novecientos setenta y cinco."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para gestionar su devolución, acceda a https://sapo.pt/faq con su número de pedido X-12340567."}, {"role": "assistant", "content": "Para gestionar su devolución, acceda a sapo punto pt barra faq con su número de pedido X uno dos tres cuatro cero cinco seis siete."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 23/07/2017 del año 1985."}, {"role": "assistant", "content": "La cuenta fue abierta el veintitrés de julio de dos mil diecisiete del año mil novecientos ochenta y cinco."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hemos detectado un error en su dirección; por favor llámenos al 641 746 630."}, {"role": "assistant", "content": "Hemos detectado un error en su dirección; por favor llámenos al seis cuatro uno siete cuatro seis seis tres cero."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para gestionar su devolución, acceda a https://ziggo.nl con su número de pedido REF-2024-0113."}, {"role": "assistant", "content": "Para gestionar su devolución, acceda a ziggo punto nl con su número de pedido REF dos cero dos cuatro cero uno uno tres."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 3 de marzo de 2002 del año 1992."}, {"role": "assistant", "content": "La cuenta fue abierta el tres de marzo de dos mil dos del año mil novecientos noventa y dos."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 14/11/2005 del año 1991."}, {"role": "assistant", "content": "La cuenta fue abierta el catorce de noviembre de dos mil cinco del año mil novecientos noventa y uno."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 17/06/1997 del año 2010."}, {"role": "assistant", "content": "La cuenta fue abierta el diecisiete de junio de mil novecientos noventa y siete del año dos mil diez."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hemos detectado un error en su dirección; por favor llámenos al 624 415 542."}, {"role": "assistant", "content": "Hemos detectado un error en su dirección; por favor llámenos al seis dos cuatro cuatro uno cinco cinco cuatro dos."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 15/07/2016 del año 2008."}, {"role": "assistant", "content": "La cuenta fue abierta el quince de julio de dos mil dieciséis del año dos mil ocho."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para gestionar su devolución, acceda a https://www.gmail.com/support con su número de pedido TK-88213."}, {"role": "assistant", "content": "Para gestionar su devolución, acceda a w w w punto gmail punto com barra support con su número de pedido TK ocho ocho dos uno tres."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hemos detectado un error en su dirección; por favor llámenos al 960 491 001."}, {"role": "assistant", "content": "Hemos detectado un error en su dirección; por favor llámenos al nueve seis cero cuatro nueve uno cero cero uno."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hemos detectado un error en su dirección; por favor llámenos al +34 644 282 375."}, {"role": "assistant", "content": "Hemos detectado un error en su dirección; por favor llámenos al más tres cuatro seis cuatro cuatro dos ocho dos tres siete cinco."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de EUR 93 en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de noventa y tres euros en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 11 de julio de 2004 del año 1966."}, {"role": "assistant", "content": "La cuenta fue abierta el once de julio de dos mil cuatro del año mil novecientos sesenta y seis."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para gestionar su devolución, acceda a www.mail.fr con su número de pedido REF-2024-0113."}, {"role": "assistant", "content": "Para gestionar su devolución, acceda a w w w punto mail punto fr con su número de pedido REF dos cero dos cuatro cero uno uno tres."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para más información sobre su envío TK-88213, visite nuestro portal en www.ziggo.nl/login."}, {"role": "assistant", "content": "Para más información sobre su envío TK ocho ocho dos uno tres, visite nuestro portal en w w w punto ziggo punto nl barra login."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para más información sobre su envío ABC1234, visite nuestro portal en www.yahoo.com/login."}, {"role": "assistant", "content": "Para más información sobre su envío ABC uno dos tres cuatro, visite nuestro portal en w w w punto yahoo punto com barra login."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de 44,90€ en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de cuarenta y cuatro euros con noventa céntimos en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 27 de enero del año 1980."}, {"role": "assistant", "content": "La cuenta fue abierta el veintisiete de enero del año mil novecientos ochenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 27/3/2022 del año 2015."}, {"role": "assistant", "content": "La cuenta fue abierta el veintisiete de marzo de dos mil veintidós del año dos mil quince."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para gestionar su devolución, acceda a https://www.yahoo.com.ph/support con su número de pedido X-12340567."}, {"role": "assistant", "content": "Para gestionar su devolución, acceda a w w w punto yahoo punto com punto ph barra support con su número de pedido X uno dos tres cuatro cero cinco seis siete."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hemos detectado un error en su dirección; por favor llámenos al +34 651 664 673."}, {"role": "assistant", "content": "Hemos detectado un error en su dirección; por favor llámenos al más tres cuatro seis cinco uno seis seis cuatro seis siete tres."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de 383,90 € en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de trescientos ochenta y tres euros con noventa céntimos en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para más información sobre su envío TK-88213, visite nuestro portal en www.scicom.com.my."}, {"role": "assistant", "content": "Para más información sobre su envío TK ocho ocho dos uno tres, visite nuestro portal en w w w punto scicom punto com punto my."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hemos detectado un error en su dirección; por favor llámenos al 666 649 932."}, {"role": "assistant", "content": "Hemos detectado un error en su dirección; por favor llámenos al seis seis seis seis cuatro nueve nueve tres dos."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para gestionar su devolución, acceda a www.company.co/login con su número de pedido A12."}, {"role": "assistant", "content": "Para gestionar su devolución, acceda a w w w punto company punto co barra login con su número de pedido A uno dos."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 12/10/2023 del año 2022."}, {"role": "assistant", "content": "La cuenta fue abierta el doce de octubre de dos mil veintitrés del año dos mil veintidós."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para gestionar su devolución, acceda a https://company.co/help con su número de pedido INV0472."}, {"role": "assistant", "content": "Para gestionar su devolución, acceda a company punto co barra help con su número de pedido INV cero cuatro siete dos."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para más información sobre su envío TK-88213, visite nuestro portal en https://www.web.de/support."}, {"role": "assistant", "content": "Para más información sobre su envío TK ocho ocho dos uno tres, visite nuestro portal en w w w punto web punto de barra support."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para más información sobre su envío PO 55219, visite nuestro portal en https://company.co/faq."}, {"role": "assistant", "content": "Para más información sobre su envío PO cinco cinco dos uno nueve, visite nuestro portal en company punto co barra faq."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para gestionar su devolución, acceda a https://www.yahoo.com/support con su número de pedido QZ8501."}, {"role": "assistant", "content": "Para gestionar su devolución, acceda a w w w punto yahoo punto com barra support con su número de pedido QZ ocho cinco cero uno."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 19 de abril del año 1973."}, {"role": "assistant", "content": "La cuenta fue abierta el diecinueve de abril del año mil novecientos setenta y tres."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para más información sobre su envío PO 55219, visite nuestro portal en https://correo.es/faq."}, {"role": "assistant", "content": "Para más información sobre su envío PO cinco cinco dos uno nueve, visite nuestro portal en correo punto es barra faq."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 11/6/2005 del año 1971."}, {"role": "assistant", "content": "La cuenta fue abierta el once de junio de dos mil cinco del año mil novecientos setenta y uno."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hemos detectado un error en su dirección; por favor llámenos al +34 663 865 828."}, {"role": "assistant", "content": "Hemos detectado un error en su dirección; por favor llámenos al más tres cuatro seis seis tres ocho seis cinco ocho dos ocho."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hemos detectado un error en su dirección; por favor llámenos al 743 529 133."}, {"role": "assistant", "content": "Hemos detectado un error en su dirección; por favor llámenos al siete cuatro tres cinco dos nueve uno tres tres."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para más información sobre su envío TK-88213, visite nuestro portal en www.web.de/login."}, {"role": "assistant", "content": "Para más información sobre su envío TK ocho ocho dos uno tres, visite nuestro portal en w w w punto web punto de barra login."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de 22,26€ en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de veintidós euros con veintiséis céntimos en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de 890€ en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de ochocientos noventa euros en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hemos detectado un error en su dirección; por favor llámenos al 916 469 644."}, {"role": "assistant", "content": "Hemos detectado un error en su dirección; por favor llámenos al nueve uno seis cuatro seis nueve seis cuatro cuatro."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de 42,90€ en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de cuarenta y dos euros con noventa céntimos en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para gestionar su devolución, acceda a https://www.posta.it/support con su número de pedido AK6120."}, {"role": "assistant", "content": "Para gestionar su devolución, acceda a w w w punto posta punto it barra support con su número de pedido AK seis uno dos cero."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de 7.754,90 € en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de siete mil setecientos cincuenta y cuatro euros con noventa céntimos en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para más información sobre su envío REF-2024-0113, visite nuestro portal en https://www.example.com/support."}, {"role": "assistant", "content": "Para más información sobre su envío REF dos cero dos cuatro cero uno uno tres, visite nuestro portal en w w w punto example punto com barra support."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hemos detectado un error en su dirección; por favor llámenos al 638 058 065."}, {"role": "assistant", "content": "Hemos detectado un error en su dirección; por favor llámenos al seis tres ocho cero cinco ocho cero seis cinco."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para más información sobre su envío B2-14, visite nuestro portal en https://sapo.pt/faq."}, {"role": "assistant", "content": "Para más información sobre su envío B dos uno cuatro, visite nuestro portal en sapo punto pt barra faq."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 7 de diciembre de 2020 del año 1970."}, {"role": "assistant", "content": "La cuenta fue abierta el siete de diciembre de dos mil veinte del año mil novecientos setenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para más información sobre su envío X-12340567, visite nuestro portal en https://web.de."}, {"role": "assistant", "content": "Para más información sobre su envío X uno dos tres cuatro cero cinco seis siete, visite nuestro portal en web punto de."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hemos detectado un error en su dirección; por favor llámenos al 754 157 694."}, {"role": "assistant", "content": "Hemos detectado un error en su dirección; por favor llámenos al siete cinco cuatro uno cinco siete seis nueve cuatro."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para más información sobre su envío QZ8501, visite nuestro portal en https://www.outlook.sa/support."}, {"role": "assistant", "content": "Para más información sobre su envío QZ ocho cinco cero uno, visite nuestro portal en w w w punto outlook punto sa barra support."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para más información sobre su envío WXY 1234, visite nuestro portal en https://outlook.sa."}, {"role": "assistant", "content": "Para más información sobre su envío WXY uno dos tres cuatro, visite nuestro portal en outlook punto sa."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hemos detectado un error en su dirección; por favor llámenos al 742 371 372."}, {"role": "assistant", "content": "Hemos detectado un error en su dirección; por favor llámenos al siete cuatro dos tres siete uno tres siete dos."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para gestionar su devolución, acceda a https://company.co con su número de pedido MH370."}, {"role": "assistant", "content": "Para gestionar su devolución, acceda a company punto co con su número de pedido MH tres siete cero."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de EUR 14 en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de catorce euros en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para gestionar su devolución, acceda a https://onet.pl/help con su número de pedido MH370."}, {"role": "assistant", "content": "Para gestionar su devolución, acceda a onet punto pl barra help con su número de pedido MH tres siete cero."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para más información sobre su envío X-12340567, visite nuestro portal en https://mail.fr."}, {"role": "assistant", "content": "Para más información sobre su envío X uno dos tres cuatro cero cinco seis siete, visite nuestro portal en mail punto fr."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 03/04/1995 del año 1966."}, {"role": "assistant", "content": "La cuenta fue abierta el tres de abril de mil novecientos noventa y cinco del año mil novecientos sesenta y seis."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de 55€ en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de cincuenta y cinco euros en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para más información sobre su envío PO 55219, visite nuestro portal en www.correo.es."}, {"role": "assistant", "content": "Para más información sobre su envío PO cinco cinco dos uno nueve, visite nuestro portal en w w w punto correo punto es."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hemos detectado un error en su dirección; por favor llámenos al +34 614 253 525."}, {"role": "assistant", "content": "Hemos detectado un error en su dirección; por favor llámenos al más tres cuatro seis uno cuatro dos cinco tres cinco dos cinco."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de EUR 596 en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de quinientos noventa y seis euros en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para gestionar su devolución, acceda a www.gmail.co.id/login con su número de pedido X-12340567."}, {"role": "assistant", "content": "Para gestionar su devolución, acceda a w w w punto gmail punto co punto id barra login con su número de pedido X uno dos tres cuatro cero cinco seis siete."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 12/11/2014 del año 2010."}, {"role": "assistant", "content": "La cuenta fue abierta el doce de noviembre de dos mil catorce del año dos mil diez."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 03/02/2025 del año 2031."}, {"role": "assistant", "content": "La cuenta fue abierta el tres de febrero de dos mil veinticinco del año dos mil treinta y uno."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 06/11/2024 del año 2031."}, {"role": "assistant", "content": "La cuenta fue abierta el seis de noviembre de dos mil veinticuatro del año dos mil treinta y uno."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para gestionar su devolución, acceda a https://www.example.com/support con su número de pedido QZ8501."}, {"role": "assistant", "content": "Para gestionar su devolución, acceda a w w w punto example punto com barra support con su número de pedido QZ ocho cinco cero uno."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de EUR 77.746,90 en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de setenta y siete mil setecientos cuarenta y seis euros con noventa céntimos en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La cuenta fue abierta el 14 de junio del año 2020."}, {"role": "assistant", "content": "La cuenta fue abierta el catorce de junio del año dos mil veinte."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hemos detectado un error en su dirección; por favor llámenos al 749 322 085."}, {"role": "assistant", "content": "Hemos detectado un error en su dirección; por favor llámenos al siete cuatro nueve tres dos dos cero ocho cinco."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para gestionar su devolución, acceda a https://onet.pl con su número de pedido PO 55219."}, {"role": "assistant", "content": "Para gestionar su devolución, acceda a onet punto pl con su número de pedido PO cinco cinco dos uno nueve."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para gestionar su devolución, acceda a www.mail.fr/login con su número de pedido QZ8501."}, {"role": "assistant", "content": "Para gestionar su devolución, acceda a w w w punto mail punto fr barra login con su número de pedido QZ ocho cinco cero uno."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para más información sobre su envío PO 55219, visite nuestro portal en https://www.sltnet.lk/support."}, {"role": "assistant", "content": "Para más información sobre su envío PO cinco cinco dos uno nueve, visite nuestro portal en w w w punto sltnet punto lk barra support."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hemos detectado un error en su dirección; por favor llámenos al +34 619 714 611."}, {"role": "assistant", "content": "Hemos detectado un error en su dirección; por favor llámenos al más tres cuatro seis uno nueve siete uno cuatro seis uno uno."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para más información sobre su envío INV0472, visite nuestro portal en https://bank.my/help."}, {"role": "assistant", "content": "Para más información sobre su envío INV cero cuatro siete dos, visite nuestro portal en bank punto my barra help."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para más información sobre su envío MH370, visite nuestro portal en https://www.yahoo.com/support."}, {"role": "assistant", "content": "Para más información sobre su envío MH tres siete cero, visite nuestro portal en w w w punto yahoo punto com barra support."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para gestionar su devolución, acceda a https://gmail.co.id con su número de pedido A12."}, {"role": "assistant", "content": "Para gestionar su devolución, acceda a gmail punto co punto id con su número de pedido A uno dos."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para más información sobre su envío INV0472, visite nuestro portal en https://gmail.com."}, {"role": "assistant", "content": "Para más información sobre su envío INV cero cuatro siete dos, visite nuestro portal en gmail punto com."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para más información sobre su envío WXY 1234, visite nuestro portal en https://web.de/help."}, {"role": "assistant", "content": "Para más información sobre su envío WXY uno dos tres cuatro, visite nuestro portal en web punto de barra help."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hola, tienes un saldo disponible de 628,90€ en tu cuenta."}, {"role": "assistant", "content": "Hola, tienes un saldo disponible de seiscientos veintiocho euros con noventa céntimos en tu cuenta."}], "lang": "es", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Guten Tag, wir konnten die Lieferung am 10. Versuch leider nicht zustellen."}, {"role": "assistant", "content": "Guten Tag, wir konnten die Lieferung am zehnte Versuch leider nicht zustellen."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Guten Tag, wir konnten die Lieferung am 12. Versuch leider nicht zustellen."}, {"role": "assistant", "content": "Guten Tag, wir konnten die Lieferung am zwölfte Versuch leider nicht zustellen."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer INV0472 wurde am 22. November 1998 erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer INV null vier sieben zwei wurde am zweiundzwanzigsten November neunzehnhundertachtundneunzig erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Guten Tag, wir konnten die Lieferung am 5. Versuch leider nicht zustellen."}, {"role": "assistant", "content": "Guten Tag, wir konnten die Lieferung am fünfte Versuch leider nicht zustellen."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer MH370 wurde am 3.3.2000 erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer MH drei sieben null wurde am dritten März zweitausend erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer ABC1234 wurde am 13. Dezember erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer ABC eins zwei drei vier wurde am dreizehnten Dezember erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer AK6120 wird heute zwischen 29 - 36 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer AK sechs eins zwei null wird heute zwischen neunundzwanzig bis sechsunddreißig Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 211 kg pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von zweihundertelf Kilogramm pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Guten Tag, wir konnten die Lieferung am 2. Versuch leider nicht zustellen."}, {"role": "assistant", "content": "Guten Tag, wir konnten die Lieferung am zweite Versuch leider nicht zustellen."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Prüfung findet in 20–24 Tagen statt."}, {"role": "assistant", "content": "Die Prüfung findet in zwanzig bis vierundzwanzig Tagen statt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Prüfung findet in 7 - 8 Tagen statt."}, {"role": "assistant", "content": "Die Prüfung findet in sieben bis acht Tagen statt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 5l pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von fünf Liter pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Guten Tag, wir konnten die Lieferung am 8. Versuch leider nicht zustellen."}, {"role": "assistant", "content": "Guten Tag, wir konnten die Lieferung am achte Versuch leider nicht zustellen."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer QZ8501 wurde am 24. Mai 2016 erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer QZ acht fünf null eins wurde am vierundzwanzigsten Mai zweitausendsechzehn erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Guten Tag, wir konnten die Lieferung am 28. Versuch leider nicht zustellen."}, {"role": "assistant", "content": "Guten Tag, wir konnten die Lieferung am achtundzwanzigste Versuch leider nicht zustellen."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer REF-2024-0113 wird heute zwischen 12–22 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer REF zwei null zwei vier null eins eins drei wird heute zwischen zwölf bis zweiundzwanzig Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer MH370 wird heute zwischen 5 - 15 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer MH drei sieben null wird heute zwischen fünf bis fünfzehn Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer B2-14 wird heute zwischen 13 - 20 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer B zwei eins vier wird heute zwischen dreizehn bis zwanzig Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Prüfung findet in 3-9 Tagen statt."}, {"role": "assistant", "content": "Die Prüfung findet in drei bis neun Tagen statt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Guten Tag, wir konnten die Lieferung am 31. Versuch leider nicht zustellen."}, {"role": "assistant", "content": "Guten Tag, wir konnten die Lieferung am einunddreißigste Versuch leider nicht zustellen."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer X-12340567 wird heute zwischen 5 - 8 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer X eins zwei drei vier null fünf sechs sieben wird heute zwischen fünf bis acht Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 1,1 Liter pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von eins Komma eins Liter pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer X-12340567 wurde am 10.12.2026 erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer X eins zwei drei vier null fünf sechs sieben wurde am zehnten Dezember zweitausendsechsundzwanzig erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer REF-2024-0113 wurde am 11. April erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer REF zwei null zwei vier null eins eins drei wurde am elften April erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer A12 wurde am 26.7.2021 erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer A eins zwei wurde am sechsundzwanzigsten Juli zweitausendeinundzwanzig erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 315,1 cm pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von dreihundertfünfzehn Komma eins Zentimeter pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer A12 wurde am 9.10.2012 erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer A eins zwei wurde am neunten Oktober zweitausendzwölf erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer AK6120 wird heute zwischen 20 - 23 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer AK sechs eins zwei null wird heute zwischen zwanzig bis dreiundzwanzig Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer MH370 wird heute zwischen 13 - 16 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer MH drei sieben null wird heute zwischen dreizehn bis sechzehn Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer WXY 1234 wurde am 23. Januar 2031 erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer WXY eins zwei drei vier wurde am dreiundzwanzigsten Januar zweitausendeinunddreißig erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer TK-88213 wird heute zwischen 19-23 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer TK acht acht zwei eins drei wird heute zwischen neunzehn bis dreiundzwanzig Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer INV0472 wird heute zwischen 29–39 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer INV null vier sieben zwei wird heute zwischen neunundzwanzig bis neununddreißig Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer WXY 1234 wird heute zwischen 17–24 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer WXY eins zwei drei vier wird heute zwischen siebzehn bis vierundzwanzig Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 1,3 °C pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von eins Komma drei Grad Celsius pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer MH370 wurde am 16.7.2017 erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer MH drei sieben null wurde am sechzehnten Juli zweitausendsiebzehn erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Prüfung findet in 12–15 Tagen statt."}, {"role": "assistant", "content": "Die Prüfung findet in zwölf bis fünfzehn Tagen statt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Prüfung findet in 29-35 Tagen statt."}, {"role": "assistant", "content": "Die Prüfung findet in neunundzwanzig bis fünfunddreißig Tagen statt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer INV0472 wurde am 23.09.2003 erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer INV null vier sieben zwei wurde am dreiundzwanzigsten September zweitausenddrei erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Prüfung findet in 11-13 Tagen statt."}, {"role": "assistant", "content": "Die Prüfung findet in elf bis dreizehn Tagen statt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer TK-88213 wurde am 20.07.2031 erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer TK acht acht zwei eins drei wurde am zwanzigsten Juli zweitausendeinunddreißig erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Prüfung findet in 9-10 Tagen statt."}, {"role": "assistant", "content": "Die Prüfung findet in neun bis zehn Tagen statt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 327cm pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von dreihundertsiebenundzwanzig Zentimeter pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Prüfung findet in 17-24 Tagen statt."}, {"role": "assistant", "content": "Die Prüfung findet in siebzehn bis vierundzwanzig Tagen statt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 1km pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von ein Kilometer pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Guten Tag, wir konnten die Lieferung am 29. Versuch leider nicht zustellen."}, {"role": "assistant", "content": "Guten Tag, wir konnten die Lieferung am neunundzwanzigste Versuch leider nicht zustellen."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer QZ8501 wurde am 12.3.2019 erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer QZ acht fünf null eins wurde am zwölften März zweitausendneunzehn erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 1ml pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von ein Milliliter pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Prüfung findet in 22-30 Tagen statt."}, {"role": "assistant", "content": "Die Prüfung findet in zweiundzwanzig bis dreißig Tagen statt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Prüfung findet in 9 - 15 Tagen statt."}, {"role": "assistant", "content": "Die Prüfung findet in neun bis fünfzehn Tagen statt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 282 l pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von zweihundertzweiundachtzig Liter pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer WXY 1234 wird heute zwischen 16-21 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer WXY eins zwei drei vier wird heute zwischen sechzehn bis einundzwanzig Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer MH370 wurde am 10. August erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer MH drei sieben null wurde am zehnten August erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Guten Tag, wir konnten die Lieferung am 13. Versuch leider nicht zustellen."}, {"role": "assistant", "content": "Guten Tag, wir konnten die Lieferung am dreizehnte Versuch leider nicht zustellen."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer TK-88213 wird heute zwischen 3-9 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer TK acht acht zwei eins drei wird heute zwischen drei bis neun Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 350 ml pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von dreihundertfünfzig Milliliter pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 1,6Liter pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von eins Komma sechs Liter pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 4,4kg pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von vier Komma vier Kilogramm pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Guten Tag, wir konnten die Lieferung am 30. Versuch leider nicht zustellen."}, {"role": "assistant", "content": "Guten Tag, wir konnten die Lieferung am dreißigste Versuch leider nicht zustellen."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Guten Tag, wir konnten die Lieferung am 3. Versuch leider nicht zustellen."}, {"role": "assistant", "content": "Guten Tag, wir konnten die Lieferung am dritte Versuch leider nicht zustellen."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Prüfung findet in 2-7 Tagen statt."}, {"role": "assistant", "content": "Die Prüfung findet in zwei bis sieben Tagen statt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer MH370 wurde am 11. Juli erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer MH drei sieben null wurde am elften Juli erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer X-12340567 wurde am 5.12.2027 erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer X eins zwei drei vier null fünf sechs sieben wurde am fünften Dezember zweitausendsiebenundzwanzig erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer AK6120 wurde am 07.11.2011 erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer AK sechs eins zwei null wurde am siebten November zweitausendelf erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer TK-88213 wird heute zwischen 9–11 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer TK acht acht zwei eins drei wird heute zwischen neun bis elf Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer QZ8501 wird heute zwischen 20-29 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer QZ acht fünf null eins wird heute zwischen zwanzig bis neunundzwanzig Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Guten Tag, wir konnten die Lieferung am 26. Versuch leider nicht zustellen."}, {"role": "assistant", "content": "Guten Tag, wir konnten die Lieferung am sechsundzwanzigste Versuch leider nicht zustellen."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer MH370 wurde am 5.4.2010 erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer MH drei sieben null wurde am fünften April zweitausendzehn erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Prüfung findet in 13–15 Tagen statt."}, {"role": "assistant", "content": "Die Prüfung findet in dreizehn bis fünfzehn Tagen statt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Prüfung findet in 29–34 Tagen statt."}, {"role": "assistant", "content": "Die Prüfung findet in neunundzwanzig bis vierunddreißig Tagen statt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Prüfung findet in 2–7 Tagen statt."}, {"role": "assistant", "content": "Die Prüfung findet in zwei bis sieben Tagen statt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Guten Tag, wir konnten die Lieferung am 17. Versuch leider nicht zustellen."}, {"role": "assistant", "content": "Guten Tag, wir konnten die Lieferung am siebzehnte Versuch leider nicht zustellen."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Prüfung findet in 25–34 Tagen statt."}, {"role": "assistant", "content": "Die Prüfung findet in fünfundzwanzig bis vierunddreißig Tagen statt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Guten Tag, wir konnten die Lieferung am 6. Versuch leider nicht zustellen."}, {"role": "assistant", "content": "Guten Tag, wir konnten die Lieferung am sechste Versuch leider nicht zustellen."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer B2-14 wird heute zwischen 29–34 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer B zwei eins vier wird heute zwischen neunundzwanzig bis vierunddreißig Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 1 l pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von ein Liter pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer MH370 wird heute zwischen 22-26 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer MH drei sieben null wird heute zwischen zweiundzwanzig bis sechsundzwanzig Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Prüfung findet in 9–10 Tagen statt."}, {"role": "assistant", "content": "Die Prüfung findet in neun bis zehn Tagen statt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Prüfung findet in 23 - 26 Tagen statt."}, {"role": "assistant", "content": "Die Prüfung findet in dreiundzwanzig bis sechsundzwanzig Tagen statt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer REF-2024-0113 wird heute zwischen 20–29 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer REF zwei null zwei vier null eins eins drei wird heute zwischen zwanzig bis neunundzwanzig Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer QZ8501 wurde am 1. Juli erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer QZ acht fünf null eins wurde am ersten Juli erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer AK6120 wurde am 19. Juni erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer AK sechs eins zwei null wurde am neunzehnten Juni erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer REF-2024-0113 wird heute zwischen 17 - 27 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer REF zwei null zwei vier null eins eins drei wird heute zwischen siebzehn bis siebenundzwanzig Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Prüfung findet in 22-31 Tagen statt."}, {"role": "assistant", "content": "Die Prüfung findet in zweiundzwanzig bis einunddreißig Tagen statt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer ABC1234 wird heute zwischen 11–19 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer ABC eins zwei drei vier wird heute zwischen elf bis neunzehn Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Guten Tag, wir konnten die Lieferung am 7. Versuch leider nicht zustellen."}, {"role": "assistant", "content": "Guten Tag, wir konnten die Lieferung am siebte Versuch leider nicht zustellen."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Guten Tag, wir konnten die Lieferung am 27. Versuch leider nicht zustellen."}, {"role": "assistant", "content": "Guten Tag, wir konnten die Lieferung am siebenundzwanzigste Versuch leider nicht zustellen."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Guten Tag, wir konnten die Lieferung am 20. Versuch leider nicht zustellen."}, {"role": "assistant", "content": "Guten Tag, wir konnten die Lieferung am zwanzigste Versuch leider nicht zustellen."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 3 g pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von drei Gramm pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Guten Tag, wir konnten die Lieferung am 16. Versuch leider nicht zustellen."}, {"role": "assistant", "content": "Guten Tag, wir konnten die Lieferung am sechzehnte Versuch leider nicht zustellen."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Prüfung findet in 12 - 15 Tagen statt."}, {"role": "assistant", "content": "Die Prüfung findet in zwölf bis fünfzehn Tagen statt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer REF-2024-0113 wurde am 23. Januar erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer REF zwei null zwei vier null eins eins drei wurde am dreiundzwanzigsten Januar erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer INV0472 wird heute zwischen 8-10 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer INV null vier sieben zwei wird heute zwischen acht bis zehn Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 132,7g pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von einhundertzweiunddreißig Komma sieben Gramm pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Prüfung findet in 30–37 Tagen statt."}, {"role": "assistant", "content": "Die Prüfung findet in dreißig bis siebenunddreißig Tagen statt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 232 Liter pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von zweihundertzweiunddreißig Liter pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 5,1kg pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von fünf Komma eins Kilogramm pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 1 m pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von ein Meter pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer TK-88213 wird heute zwischen 21-30 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer TK acht acht zwei eins drei wird heute zwischen einundzwanzig bis dreißig Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer WXY 1234 wurde am 9. Oktober erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer WXY eins zwei drei vier wurde am neunten Oktober erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 9 m pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von neun Meter pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer B2-14 wurde am 14. September 2003 erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer B zwei eins vier wurde am vierzehnten September zweitausenddrei erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer A12 wird heute zwischen 3 - 5 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer A eins zwei wird heute zwischen drei bis fünf Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer REF-2024-0113 wurde am 12.07.2025 erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer REF zwei null zwei vier null eins eins drei wurde am zwölften Juli zweitausendfünfundzwanzig erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer WXY 1234 wurde am 26. November erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer WXY eins zwei drei vier wurde am sechsundzwanzigsten November erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 1 cm pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von ein Zentimeter pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 108,1Liter pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von einhundertacht Komma eins Liter pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer INV0472 wird heute zwischen 13 - 20 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer INV null vier sieben zwei wird heute zwischen dreizehn bis zwanzig Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Guten Tag, wir konnten die Lieferung am 14. Versuch leider nicht zustellen."}, {"role": "assistant", "content": "Guten Tag, wir konnten die Lieferung am vierzehnte Versuch leider nicht zustellen."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Prüfung findet in 13-16 Tagen statt."}, {"role": "assistant", "content": "Die Prüfung findet in dreizehn bis sechzehn Tagen statt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 377,8 kg pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von dreihundertsiebenundsiebzig Komma acht Kilogramm pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer AK6120 wurde am 14.12.1998 erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer AK sechs eins zwei null wurde am vierzehnten Dezember neunzehnhundertachtundneunzig erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer X-12340567 wird heute zwischen 30–32 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer X eins zwei drei vier null fünf sechs sieben wird heute zwischen dreißig bis zweiunddreißig Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer A12 wurde am 25. Februar 2031 erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer A eins zwei wurde am fünfundzwanzigsten Februar zweitausendeinunddreißig erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Prüfung findet in 24-31 Tagen statt."}, {"role": "assistant", "content": "Die Prüfung findet in vierundzwanzig bis einunddreißig Tagen statt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer ABC1234 wurde am 18. April 2009 erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer ABC eins zwei drei vier wurde am achtzehnten April zweitausendneun erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 425,3 ml pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von vierhundertfünfundzwanzig Komma drei Milliliter pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 2 kg pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von zwei Kilogramm pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 1 Liter pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von ein Liter pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer MH370 wurde am 17. Juni 1996 erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer MH drei sieben null wurde am siebzehnten Juni neunzehnhundertsechsundneunzig erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer MH370 wird heute zwischen 4-11 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer MH drei sieben null wird heute zwischen vier bis elf Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Prüfung findet in 22-27 Tagen statt."}, {"role": "assistant", "content": "Die Prüfung findet in zweiundzwanzig bis siebenundzwanzig Tagen statt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer PO 55219 wird heute zwischen 10–20 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer PO fünf fünf zwei eins neun wird heute zwischen zehn bis zwanzig Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Prüfung findet in 26-35 Tagen statt."}, {"role": "assistant", "content": "Die Prüfung findet in sechsundzwanzig bis fünfunddreißig Tagen statt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 176 Grad pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von einhundertsechsundsiebzig Grad Celsius pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer ABC1234 wird heute zwischen 22–23 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer ABC eins zwei drei vier wird heute zwischen zweiundzwanzig bis dreiundzwanzig Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 237m pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von zweihundertsiebenunddreißig Meter pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer A12 wird heute zwischen 19–24 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer A eins zwei wird heute zwischen neunzehn bis vierundzwanzig Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer REF-2024-0113 wird heute zwischen 11-21 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer REF zwei null zwei vier null eins eins drei wird heute zwischen elf bis einundzwanzig Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer MH370 wird heute zwischen 19-27 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer MH drei sieben null wird heute zwischen neunzehn bis siebenundzwanzig Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Guten Tag, wir konnten die Lieferung am 25. Versuch leider nicht zustellen."}, {"role": "assistant", "content": "Guten Tag, wir konnten die Lieferung am fünfundzwanzigste Versuch leider nicht zustellen."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer PO 55219 wird heute zwischen 15–20 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer PO fünf fünf zwei eins neun wird heute zwischen fünfzehn bis zwanzig Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 1 Grad pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von ein Grad Celsius pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer QZ8501 wurde am 16.05.2028 erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer QZ acht fünf null eins wurde am sechzehnten Mai zweitausendachtundzwanzig erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer B2-14 wurde am 8.9.2004 erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer B zwei eins vier wurde am achten September zweitausendvier erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer QZ8501 wurde am 11.11.2030 erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer QZ acht fünf null eins wurde am elften November zweitausenddreißig erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 1,1ml pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von eins Komma eins Milliliter pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer PO 55219 wird heute zwischen 6–9 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer PO fünf fünf zwei eins neun wird heute zwischen sechs bis neun Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Prüfung findet in 11 - 17 Tagen statt."}, {"role": "assistant", "content": "Die Prüfung findet in elf bis siebzehn Tagen statt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer TK-88213 wird heute zwischen 24-31 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer TK acht acht zwei eins drei wird heute zwischen vierundzwanzig bis einunddreißig Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 1m pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von ein Meter pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer PO 55219 wird heute zwischen 11 - 14 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer PO fünf fünf zwei eins neun wird heute zwischen elf bis vierzehn Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 123Liter pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von einhundertdreiundzwanzig Liter pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Hallo! Ihr Paket mit der Sendungsnummer INV0472 wird heute zwischen 17 - 24 Uhr zugestellt."}, {"role": "assistant", "content": "Hallo! Ihr Paket mit der Sendungsnummer INV null vier sieben zwei wird heute zwischen siebzehn bis vierundzwanzig Uhr zugestellt."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Paket beinhaltet ein Datenvolumen von 275,4 cm pro Monat."}, {"role": "assistant", "content": "Ihr Paket beinhaltet ein Datenvolumen von zweihundertfünfundsiebzig Komma vier Zentimeter pro Monat."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Antrag mit der Nummer ABC1234 wurde am 2. August erfolgreich eingereicht."}, {"role": "assistant", "content": "Ihr Antrag mit der Nummer ABC eins zwei drei vier wurde am zweiten August erfolgreich eingereicht."}], "lang": "de", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La tua pratica numero AK6120 è stata elaborata con successo."}, {"role": "assistant", "content": "La tua pratica numero AK sei uno due zero è stata elaborata con successo."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il dosaggio consigliato per il farmaco è di 28,94 mg per dose."}, {"role": "assistant", "content": "Il dosaggio consigliato per il farmaco è di ventotto virgola nove quattro mg per dose."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il dosaggio consigliato per il farmaco è di 31,4 mg per dose."}, {"role": "assistant", "content": "Il dosaggio consigliato per il farmaco è di trentuno virgola quattro mg per dose."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il dosaggio consigliato per il farmaco è di 33,5 mg per dose."}, {"role": "assistant", "content": "Il dosaggio consigliato per il farmaco è di trentatré virgola cinque mg per dose."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il punteggio minimo richiesto per superare l'esame è di 78,97 su 5."}, {"role": "assistant", "content": "Il punteggio minimo richiesto per superare l'esame è di settantotto virgola nove sette su cinque."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La tua pratica numero REF-2024-0113 è stata elaborata con successo."}, {"role": "assistant", "content": "La tua pratica numero REF due zero due quattro zero uno uno tre è stata elaborata con successo."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il dosaggio consigliato per il farmaco è di 64,87 mg per dose."}, {"role": "assistant", "content": "Il dosaggio consigliato per il farmaco è di sessantaquattro virgola otto sette mg per dose."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ti informiamo che il tuo abbonamento scadrà il 24/3/2019 alle ore 6.45."}, {"role": "assistant", "content": "Ti informiamo che il tuo abbonamento scadrà il ventiquattro marzo duemiladiciannove alle ore sei e quarantacinque."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il dosaggio consigliato per il farmaco è di 22,08 mg per dose."}, {"role": "assistant", "content": "Il dosaggio consigliato per il farmaco è di ventidue virgola zero otto mg per dose."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ti informiamo che il tuo abbonamento scadrà il 14 maggio 2023 alle ore 5.15."}, {"role": "assistant", "content": "Ti informiamo che il tuo abbonamento scadrà il quattordici maggio duemilaventitre alle ore cinque e quindici."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il punteggio minimo richiesto per superare l'esame è di 59,3 su 95."}, {"role": "assistant", "content": "Il punteggio minimo richiesto per superare l'esame è di cinquantanove virgola tre su novantacinque."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ti informiamo che il tuo abbonamento scadrà il 17 gennaio alle ore 22:20."}, {"role": "assistant", "content": "Ti informiamo che il tuo abbonamento scadrà il diciassette gennaio alle ore ventidue e venti."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il punteggio minimo richiesto per superare l'esame è di 73,3 su 50."}, {"role": "assistant", "content": "Il punteggio minimo richiesto per superare l'esame è di settantatré virgola tre su cinquanta."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il dosaggio consigliato per il farmaco è di 58,9 mg per dose."}, {"role": "assistant", "content": "Il dosaggio consigliato per il farmaco è di cinquantotto virgola nove mg per dose."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La tua pratica numero B2-14 è stata elaborata con successo."}, {"role": "assistant", "content": "La tua pratica numero B due uno quattro è stata elaborata con successo."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura TK-88213 è fissata per il 6 maggio."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura TK otto otto due uno tre è fissata per il sei maggio."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La tua pratica numero A12 è stata elaborata con successo."}, {"role": "assistant", "content": "La tua pratica numero A uno due è stata elaborata con successo."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura WXY 1234 è fissata per il 24/12/2001."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura WXY uno due tre quattro è fissata per il ventiquattro dicembre duemilauno."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura A12 è fissata per il 9 maggio."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura A uno due è fissata per il nove maggio."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il dosaggio consigliato per il farmaco è di 39,7 mg per dose."}, {"role": "assistant", "content": "Il dosaggio consigliato per il farmaco è di trentanove virgola sette mg per dose."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il dosaggio consigliato per il farmaco è di 93,6 mg per dose."}, {"role": "assistant", "content": "Il dosaggio consigliato per il farmaco è di novantatré virgola sei mg per dose."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il punteggio minimo richiesto per superare l'esame è di 57,26 su 5."}, {"role": "assistant", "content": "Il punteggio minimo richiesto per superare l'esame è di cinquantasette virgola due sei su cinque."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura REF-2024-0113 è fissata per il 27/1/2027."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura REF due zero due quattro zero uno uno tre è fissata per il ventisette gennaio duemilaventisette."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura MH370 è fissata per il 8 ottobre."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura MH tre sette zero è fissata per il otto ottobre."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura TK-88213 è fissata per il 12/4/2028."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura TK otto otto due uno tre è fissata per il dodici aprile duemilaventotto."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il punteggio minimo richiesto per superare l'esame è di 12,7 su 960."}, {"role": "assistant", "content": "Il punteggio minimo richiesto per superare l'esame è di dodici virgola sette su novecentosessanta."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura ABC1234 è fissata per il 28/09/2002."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura ABC uno due tre quattro è fissata per il ventotto settembre duemiladue."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il punteggio minimo richiesto per superare l'esame è di 52,47 su 8."}, {"role": "assistant", "content": "Il punteggio minimo richiesto per superare l'esame è di cinquantadue virgola quattro sette su otto."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura INV0472 è fissata per il 13/7/1997."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura INV zero quattro sette due è fissata per il tredici luglio millenovecentonovantasette."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ti informiamo che il tuo abbonamento scadrà il 9 gennaio alle ore 14:10."}, {"role": "assistant", "content": "Ti informiamo che il tuo abbonamento scadrà il nove gennaio alle ore quattordici e dieci."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il punteggio minimo richiesto per superare l'esame è di 24,5 su 9."}, {"role": "assistant", "content": "Il punteggio minimo richiesto per superare l'esame è di ventiquattro virgola cinque su nove."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il punteggio minimo richiesto per superare l'esame è di 6,03 su 3."}, {"role": "assistant", "content": "Il punteggio minimo richiesto per superare l'esame è di sei virgola zero tre su tre."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ti informiamo che il tuo abbonamento scadrà il 28/09/1996 alle ore 14.45."}, {"role": "assistant", "content": "Ti informiamo che il tuo abbonamento scadrà il ventotto settembre millenovecentonovantasei alle ore quattordici e quarantacinque."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La tua pratica numero QZ8501 è stata elaborata con successo."}, {"role": "assistant", "content": "La tua pratica numero QZ otto cinque zero uno è stata elaborata con successo."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura MH370 è fissata per il 08/03/2011."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura MH tre sette zero è fissata per il otto marzo duemilaundici."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura AK6120 è fissata per il 21/12/2031."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura AK sei uno due zero è fissata per il ventuno dicembre duemilatrentuno."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura REF-2024-0113 è fissata per il 13/09/2030."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura REF due zero due quattro zero uno uno tre è fissata per il tredici settembre duemilatrenta."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La tua pratica numero X-12340567 è stata elaborata con successo."}, {"role": "assistant", "content": "La tua pratica numero X uno due tre quattro zero cinque sei sette è stata elaborata con successo."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il punteggio minimo richiesto per superare l'esame è di 70,40 su 865."}, {"role": "assistant", "content": "Il punteggio minimo richiesto per superare l'esame è di settanta virgola quattro zero su ottocentosessantacinque."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il dosaggio consigliato per il farmaco è di 18,55 mg per dose."}, {"role": "assistant", "content": "Il dosaggio consigliato per il farmaco è di diciotto virgola cinque cinque mg per dose."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il punteggio minimo richiesto per superare l'esame è di 59,2 su 795."}, {"role": "assistant", "content": "Il punteggio minimo richiesto per superare l'esame è di cinquantanove virgola due su settecentonovantacinque."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura MH370 è fissata per il 26/04/2022."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura MH tre sette zero è fissata per il ventisei aprile duemilaventidue."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il punteggio minimo richiesto per superare l'esame è di 3,22 su 19."}, {"role": "assistant", "content": "Il punteggio minimo richiesto per superare l'esame è di tre virgola due due su diciannove."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ti informiamo che il tuo abbonamento scadrà il 23/02/2027 alle ore 15:05."}, {"role": "assistant", "content": "Ti informiamo che il tuo abbonamento scadrà il ventitré febbraio duemilaventisette alle ore quindici e cinque."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura X-12340567 è fissata per il 13 maggio 2030."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura X uno due tre quattro zero cinque sei sette è fissata per il tredici maggio duemilatrenta."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il punteggio minimo richiesto per superare l'esame è di 50,9 su 444."}, {"role": "assistant", "content": "Il punteggio minimo richiesto per superare l'esame è di cinquanta virgola nove su quattrocentoquarantaquattro."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ti informiamo che il tuo abbonamento scadrà il 4 marzo alle ore 6.20."}, {"role": "assistant", "content": "Ti informiamo che il tuo abbonamento scadrà il quattro marzo alle ore sei e venti."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura TK-88213 è fissata per il 27/11/2020."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura TK otto otto due uno tre è fissata per il ventisette novembre duemilaventi."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura QZ8501 è fissata per il 18/4/2024."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura QZ otto cinque zero uno è fissata per il diciotto aprile duemilaventiquattro."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La tua pratica numero TK-88213 è stata elaborata con successo."}, {"role": "assistant", "content": "La tua pratica numero TK otto otto due uno tre è stata elaborata con successo."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura TK-88213 è fissata per il 4 novembre 2022."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura TK otto otto due uno tre è fissata per il quattro novembre duemilaventidue."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il dosaggio consigliato per il farmaco è di 77,97 mg per dose."}, {"role": "assistant", "content": "Il dosaggio consigliato per il farmaco è di settantasette virgola nove sette mg per dose."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il dosaggio consigliato per il farmaco è di 62,58 mg per dose."}, {"role": "assistant", "content": "Il dosaggio consigliato per il farmaco è di sessantadue virgola cinque otto mg per dose."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il dosaggio consigliato per il farmaco è di 55,32 mg per dose."}, {"role": "assistant", "content": "Il dosaggio consigliato per il farmaco è di cinquantacinque virgola tre due mg per dose."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il dosaggio consigliato per il farmaco è di 57,90 mg per dose."}, {"role": "assistant", "content": "Il dosaggio consigliato per il farmaco è di cinquantasette virgola nove zero mg per dose."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il dosaggio consigliato per il farmaco è di 59,5 mg per dose."}, {"role": "assistant", "content": "Il dosaggio consigliato per il farmaco è di cinquantanove virgola cinque mg per dose."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ti informiamo che il tuo abbonamento scadrà il 16 giugno alle ore 17.10."}, {"role": "assistant", "content": "Ti informiamo che il tuo abbonamento scadrà il sedici giugno alle ore diciassette e dieci."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura REF-2024-0113 è fissata per il 20/03/1995."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura REF due zero due quattro zero uno uno tre è fissata per il venti marzo millenovecentonovantacinque."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura QZ8501 è fissata per il 14 novembre 2018."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura QZ otto cinque zero uno è fissata per il quattordici novembre duemiladiciotto."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il punteggio minimo richiesto per superare l'esame è di 79,42 su 828."}, {"role": "assistant", "content": "Il punteggio minimo richiesto per superare l'esame è di settantanove virgola quattro due su ottocentoventotto."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La tua pratica numero PO 55219 è stata elaborata con successo."}, {"role": "assistant", "content": "La tua pratica numero PO cinque cinque due uno nove è stata elaborata con successo."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ti informiamo che il tuo abbonamento scadrà il 09/08/1998 alle ore 16.00."}, {"role": "assistant", "content": "Ti informiamo che il tuo abbonamento scadrà il nove agosto millenovecentonovantotto alle ore sedici."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il dosaggio consigliato per il farmaco è di 51,7 mg per dose."}, {"role": "assistant", "content": "Il dosaggio consigliato per il farmaco è di cinquantuno virgola sette mg per dose."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura INV0472 è fissata per il 3 febbraio."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura INV zero quattro sette due è fissata per il tre febbraio."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ti informiamo che il tuo abbonamento scadrà il 13/8/2032 alle ore 3.45."}, {"role": "assistant", "content": "Ti informiamo che il tuo abbonamento scadrà il tredici agosto duemilatrentadue alle ore tre e quarantacinque."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il punteggio minimo richiesto per superare l'esame è di 11,97 su 8."}, {"role": "assistant", "content": "Il punteggio minimo richiesto per superare l'esame è di undici virgola nove sette su otto."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura AK6120 è fissata per il 24/1/2024."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura AK sei uno due zero è fissata per il ventiquattro gennaio duemilaventiquattro."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura X-12340567 è fissata per il 11/1/2021."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura X uno due tre quattro zero cinque sei sette è fissata per il undici gennaio duemilaventuno."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il dosaggio consigliato per il farmaco è di 7,85 mg per dose."}, {"role": "assistant", "content": "Il dosaggio consigliato per il farmaco è di sette virgola otto cinque mg per dose."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il punteggio minimo richiesto per superare l'esame è di 17,2 su 57."}, {"role": "assistant", "content": "Il punteggio minimo richiesto per superare l'esame è di diciassette virgola due su cinquantasette."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il dosaggio consigliato per il farmaco è di 46,31 mg per dose."}, {"role": "assistant", "content": "Il dosaggio consigliato per il farmaco è di quarantasei virgola tre uno mg per dose."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il punteggio minimo richiesto per superare l'esame è di 39,2 su 3."}, {"role": "assistant", "content": "Il punteggio minimo richiesto per superare l'esame è di trentanove virgola due su tre."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura A12 è fissata per il 15 luglio 2014."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura A uno due è fissata per il quindici luglio duemilaquattordici."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura PO 55219 è fissata per il 23/06/2031."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura PO cinque cinque due uno nove è fissata per il ventitré giugno duemilatrentuno."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il punteggio minimo richiesto per superare l'esame è di 82,52 su 703."}, {"role": "assistant", "content": "Il punteggio minimo richiesto per superare l'esame è di ottantadue virgola cinque due su settecentotré."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il dosaggio consigliato per il farmaco è di 90,94 mg per dose."}, {"role": "assistant", "content": "Il dosaggio consigliato per il farmaco è di novanta virgola nove quattro mg per dose."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ti informiamo che il tuo abbonamento scadrà il 8 maggio 2012 alle ore 13:20."}, {"role": "assistant", "content": "Ti informiamo che il tuo abbonamento scadrà il otto maggio duemiladodici alle ore tredici e venti."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il dosaggio consigliato per il farmaco è di 55,7 mg per dose."}, {"role": "assistant", "content": "Il dosaggio consigliato per il farmaco è di cinquantacinque virgola sette mg per dose."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ti informiamo che il tuo abbonamento scadrà il 28 luglio alle ore 21:05."}, {"role": "assistant", "content": "Ti informiamo che il tuo abbonamento scadrà il ventotto luglio alle ore ventuno e cinque."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il punteggio minimo richiesto per superare l'esame è di 57,6 su 8."}, {"role": "assistant", "content": "Il punteggio minimo richiesto per superare l'esame è di cinquantasette virgola sei su otto."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ti informiamo che il tuo abbonamento scadrà il 7 febbraio alle ore 17:05."}, {"role": "assistant", "content": "Ti informiamo che il tuo abbonamento scadrà il sette febbraio alle ore diciassette e cinque."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il punteggio minimo richiesto per superare l'esame è di 34,7 su 37."}, {"role": "assistant", "content": "Il punteggio minimo richiesto per superare l'esame è di trentaquattro virgola sette su trentasette."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ti informiamo che il tuo abbonamento scadrà il 06/10/2016 alle ore 1:10."}, {"role": "assistant", "content": "Ti informiamo che il tuo abbonamento scadrà il sei ottobre duemilasedici alle ore uno e dieci."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ti informiamo che il tuo abbonamento scadrà il 10/01/2021 alle ore 22.40."}, {"role": "assistant", "content": "Ti informiamo che il tuo abbonamento scadrà il dieci gennaio duemilaventuno alle ore ventidue e quaranta."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il dosaggio consigliato per il farmaco è di 45,81 mg per dose."}, {"role": "assistant", "content": "Il dosaggio consigliato per il farmaco è di quarantacinque virgola otto uno mg per dose."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura PO 55219 è fissata per il 5/1/2000."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura PO cinque cinque due uno nove è fissata per il cinque gennaio duemila."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura X-12340567 è fissata per il 27/11/2015."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura X uno due tre quattro zero cinque sei sette è fissata per il ventisette novembre duemilaquindici."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il dosaggio consigliato per il farmaco è di 30,49 mg per dose."}, {"role": "assistant", "content": "Il dosaggio consigliato per il farmaco è di trenta virgola quattro nove mg per dose."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il punteggio minimo richiesto per superare l'esame è di 86,2 su 5."}, {"role": "assistant", "content": "Il punteggio minimo richiesto per superare l'esame è di ottantasei virgola due su cinque."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il dosaggio consigliato per il farmaco è di 8,4 mg per dose."}, {"role": "assistant", "content": "Il dosaggio consigliato per il farmaco è di otto virgola quattro mg per dose."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il punteggio minimo richiesto per superare l'esame è di 88,7 su 590."}, {"role": "assistant", "content": "Il punteggio minimo richiesto per superare l'esame è di ottantotto virgola sette su cinquecentonovanta."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il punteggio minimo richiesto per superare l'esame è di 10,32 su 5."}, {"role": "assistant", "content": "Il punteggio minimo richiesto per superare l'esame è di dieci virgola tre due su cinque."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il dosaggio consigliato per il farmaco è di 90,5 mg per dose."}, {"role": "assistant", "content": "Il dosaggio consigliato per il farmaco è di novanta virgola cinque mg per dose."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ti informiamo che il tuo abbonamento scadrà il 13 agosto 2020 alle ore 19:15."}, {"role": "assistant", "content": "Ti informiamo che il tuo abbonamento scadrà il tredici agosto duemilaventi alle ore diciannove e quindici."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ti informiamo che il tuo abbonamento scadrà il 12/7/2001 alle ore 0:30."}, {"role": "assistant", "content": "Ti informiamo che il tuo abbonamento scadrà il dodici luglio duemilauno alle ore zero e trenta."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura QZ8501 è fissata per il 17/9/2021."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura QZ otto cinque zero uno è fissata per il diciassette settembre duemilaventuno."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura AK6120 è fissata per il 23 gennaio 2018."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura AK sei uno due zero è fissata per il ventitré gennaio duemiladiciotto."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il dosaggio consigliato per il farmaco è di 11,7 mg per dose."}, {"role": "assistant", "content": "Il dosaggio consigliato per il farmaco è di undici virgola sette mg per dose."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ti informiamo che il tuo abbonamento scadrà il 4 novembre 2028 alle ore 18.00."}, {"role": "assistant", "content": "Ti informiamo che il tuo abbonamento scadrà il quattro novembre duemilaventotto alle ore diciotto."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura QZ8501 è fissata per il 22/12/2015."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura QZ otto cinque zero uno è fissata per il ventidue dicembre duemilaquindici."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ti informiamo che il tuo abbonamento scadrà il 23/09/1995 alle ore 15:30."}, {"role": "assistant", "content": "Ti informiamo che il tuo abbonamento scadrà il ventitré settembre millenovecentonovantacinque alle ore quindici e trenta."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura REF-2024-0113 è fissata per il 4/11/2029."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura REF due zero due quattro zero uno uno tre è fissata per il quattro novembre duemilaventinove."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura A12 è fissata per il 4/11/1997."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura A uno due è fissata per il quattro novembre millenovecentonovantasette."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il punteggio minimo richiesto per superare l'esame è di 44,63 su 8."}, {"role": "assistant", "content": "Il punteggio minimo richiesto per superare l'esame è di quarantaquattro virgola sei tre su otto."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ti informiamo che il tuo abbonamento scadrà il 22/3/2029 alle ore 4:10."}, {"role": "assistant", "content": "Ti informiamo che il tuo abbonamento scadrà il ventidue marzo duemilaventinove alle ore quattro e dieci."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura PO 55219 è fissata per il 24 luglio."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura PO cinque cinque due uno nove è fissata per il ventiquattro luglio."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura MH370 è fissata per il 3 febbraio 2024."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura MH tre sette zero è fissata per il tre febbraio duemilaventiquattro."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ti informiamo che il tuo abbonamento scadrà il 14 maggio alle ore 11:10."}, {"role": "assistant", "content": "Ti informiamo che il tuo abbonamento scadrà il quattordici maggio alle ore undici e dieci."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il dosaggio consigliato per il farmaco è di 28,38 mg per dose."}, {"role": "assistant", "content": "Il dosaggio consigliato per il farmaco è di ventotto virgola tre otto mg per dose."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura ABC1234 è fissata per il 22 novembre."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura ABC uno due tre quattro è fissata per il ventidue novembre."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ti informiamo che il tuo abbonamento scadrà il 2 giugno alle ore 20:00."}, {"role": "assistant", "content": "Ti informiamo che il tuo abbonamento scadrà il due giugno alle ore venti."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ti informiamo che il tuo abbonamento scadrà il 12/07/2013 alle ore 22.45."}, {"role": "assistant", "content": "Ti informiamo che il tuo abbonamento scadrà il dodici luglio duemilatredici alle ore ventidue e quarantacinque."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ti informiamo che il tuo abbonamento scadrà il 08/12/2028 alle ore 19.40."}, {"role": "assistant", "content": "Ti informiamo che il tuo abbonamento scadrà il otto dicembre duemilaventotto alle ore diciannove e quaranta."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura PO 55219 è fissata per il 10 gennaio 2021."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura PO cinque cinque due uno nove è fissata per il dieci gennaio duemilaventuno."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il dosaggio consigliato per il farmaco è di 24,5 mg per dose."}, {"role": "assistant", "content": "Il dosaggio consigliato per il farmaco è di ventiquattro virgola cinque mg per dose."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il dosaggio consigliato per il farmaco è di 3,51 mg per dose."}, {"role": "assistant", "content": "Il dosaggio consigliato per il farmaco è di tre virgola cinque uno mg per dose."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura MH370 è fissata per il 4/9/2006."}, {"role": "assistant", "content": "Gentile cliente, le ricordiamo che la scadenza della sua fattura MH tre sette zero è fissata per il quattro settembre duemilasei."}], "lang": "it", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para 914 902 647."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para nove um quatro nove zero dois seis quatro sete."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Houve um atraso de 1% no tempo estimado de entrega."}, {"role": "assistant", "content": "Houve um atraso de um por cento no tempo estimado de entrega."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 11 - 16 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de onze a dezasseis dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Você tem 322 notificações pendentes no seu portal do paciente."}, {"role": "assistant", "content": "Você tem trezentos e vinte e dois notificações pendentes no seu portal do paciente."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Você tem 709 notificações pendentes no seu portal do paciente."}, {"role": "assistant", "content": "Você tem setecentos e nove notificações pendentes no seu portal do paciente."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para 21 398 2152."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para dois um três nove oito dois um cinco dois."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para 21 886 8978."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para dois um oito oito seis oito nove sete oito."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Você tem 222 notificações pendentes no seu portal do paciente."}, {"role": "assistant", "content": "Você tem duzentos e vinte e dois notificações pendentes no seu portal do paciente."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Houve um atraso de 78,1% no tempo estimado de entrega."}, {"role": "assistant", "content": "Houve um atraso de setenta e oito vírgula um por cento no tempo estimado de entrega."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 15-20 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de quinze a vinte dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Você tem 986 notificações pendentes no seu portal do paciente."}, {"role": "assistant", "content": "Você tem novecentos e oitenta e seis notificações pendentes no seu portal do paciente."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 7-14 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de sete a catorze dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Parabéns por conquistar o 22º lugar na olimpíada de matemática!"}, {"role": "assistant", "content": "Parabéns por conquistar o vigésimo segundo lugar na olimpíada de matemática!"}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Você tem 41 notificações pendentes no seu portal do paciente."}, {"role": "assistant", "content": "Você tem quarenta e um notificações pendentes no seu portal do paciente."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para +351 970 653 245."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para mais três cinco um nove sete zero seis cinco três dois quatro cinco."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para 901 925 840."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para nove zero um nove dois cinco oito quatro zero."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para 914 216 536."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para nove um quatro dois um seis cinco três seis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para +351 970 334 887."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para mais três cinco um nove sete zero três três quatro oito oito sete."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 19-28 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de dezanove a vinte e oito dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Houve um atraso de 61,2% no tempo estimado de entrega."}, {"role": "assistant", "content": "Houve um atraso de sessenta e um vírgula dois por cento no tempo estimado de entrega."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Você tem 27 notificações pendentes no seu portal do paciente."}, {"role": "assistant", "content": "Você tem vinte e sete notificações pendentes no seu portal do paciente."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 23 - 29 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de vinte e três a vinte e nove dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Houve um atraso de 33% no tempo estimado de entrega."}, {"role": "assistant", "content": "Houve um atraso de trinta e três por cento no tempo estimado de entrega."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 21-22 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de vinte e um a vinte e dois dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Houve um atraso de 93% no tempo estimado de entrega."}, {"role": "assistant", "content": "Houve um atraso de noventa e três por cento no tempo estimado de entrega."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Houve um atraso de 25% no tempo estimado de entrega."}, {"role": "assistant", "content": "Houve um atraso de vinte e cinco por cento no tempo estimado de entrega."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Houve um atraso de 95% no tempo estimado de entrega."}, {"role": "assistant", "content": "Houve um atraso de noventa e cinco por cento no tempo estimado de entrega."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para +351 982 013 316."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para mais três cinco um nove oito dois zero um três três um seis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Houve um atraso de 36% no tempo estimado de entrega."}, {"role": "assistant", "content": "Houve um atraso de trinta e seis por cento no tempo estimado de entrega."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Você tem 79 notificações pendentes no seu portal do paciente."}, {"role": "assistant", "content": "Você tem setenta e nove notificações pendentes no seu portal do paciente."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Parabéns por conquistar o 21º lugar na olimpíada de matemática!"}, {"role": "assistant", "content": "Parabéns por conquistar o vigésimo primeiro lugar na olimpíada de matemática!"}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Parabéns por conquistar o 29º lugar na olimpíada de matemática!"}, {"role": "assistant", "content": "Parabéns por conquistar o vigésimo nono lugar na olimpíada de matemática!"}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Parabéns por conquistar o 13º lugar na olimpíada de matemática!"}, {"role": "assistant", "content": "Parabéns por conquistar o décimo terceiro lugar na olimpíada de matemática!"}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Parabéns por conquistar o 23.º lugar na olimpíada de matemática!"}, {"role": "assistant", "content": "Parabéns por conquistar o vigésimo terceiro lugar na olimpíada de matemática!"}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 2–10 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de dois a dez dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Parabéns por conquistar o 18.º lugar na olimpíada de matemática!"}, {"role": "assistant", "content": "Parabéns por conquistar o décimo oitavo lugar na olimpíada de matemática!"}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 4-12 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de quatro a doze dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para +351 977 473 862."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para mais três cinco um nove sete sete quatro sete três oito seis dois."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Parabéns por conquistar o 15º lugar na olimpíada de matemática!"}, {"role": "assistant", "content": "Parabéns por conquistar o décimo quinto lugar na olimpíada de matemática!"}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 11-12 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de onze a doze dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 29 - 35 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de vinte e nove a trinta e cinco dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Houve um atraso de 18% no tempo estimado de entrega."}, {"role": "assistant", "content": "Houve um atraso de dezoito por cento no tempo estimado de entrega."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para 976 089 321."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para nove sete seis zero oito nove três dois um."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Houve um atraso de 90% no tempo estimado de entrega."}, {"role": "assistant", "content": "Houve um atraso de noventa por cento no tempo estimado de entrega."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para 21 758 5009."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para dois um sete cinco oito cinco zero zero nove."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para 21 498 0032."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para dois um quatro nove oito zero zero três dois."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 20-25 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de vinte a vinte e cinco dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 26-28 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de vinte e seis a vinte e oito dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 9-17 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de nove a dezassete dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Você tem 58 notificações pendentes no seu portal do paciente."}, {"role": "assistant", "content": "Você tem cinquenta e oito notificações pendentes no seu portal do paciente."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 18-26 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de dezoito a vinte e seis dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Parabéns por conquistar o 4º lugar na olimpíada de matemática!"}, {"role": "assistant", "content": "Parabéns por conquistar o quarto lugar na olimpíada de matemática!"}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Parabéns por conquistar o 12.º lugar na olimpíada de matemática!"}, {"role": "assistant", "content": "Parabéns por conquistar o décimo segundo lugar na olimpíada de matemática!"}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para 21 807 9899."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para dois um oito zero sete nove oito nove nove."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Houve um atraso de 52,1% no tempo estimado de entrega."}, {"role": "assistant", "content": "Houve um atraso de cinquenta e dois vírgula um por cento no tempo estimado de entrega."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Parabéns por conquistar o 6.º lugar na olimpíada de matemática!"}, {"role": "assistant", "content": "Parabéns por conquistar o sexto lugar na olimpíada de matemática!"}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Você tem 22 notificações pendentes no seu portal do paciente."}, {"role": "assistant", "content": "Você tem vinte e dois notificações pendentes no seu portal do paciente."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Você tem 5 notificações pendentes no seu portal do paciente."}, {"role": "assistant", "content": "Você tem cinco notificações pendentes no seu portal do paciente."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Você tem 2 notificações pendentes no seu portal do paciente."}, {"role": "assistant", "content": "Você tem dois notificações pendentes no seu portal do paciente."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 24-25 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de vinte e quatro a vinte e cinco dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para 901 663 437."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para nove zero um seis seis três quatro três sete."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para 21 257 8211."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para dois um dois cinco sete oito dois um um."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para 942 379 200."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para nove quatro dois três sete nove dois zero zero."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 20-21 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de vinte a vinte e um dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para +351 922 676 123."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para mais três cinco um nove dois dois seis sete seis um dois três."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Parabéns por conquistar o 17º lugar na olimpíada de matemática!"}, {"role": "assistant", "content": "Parabéns por conquistar o décimo sétimo lugar na olimpíada de matemática!"}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Parabéns por conquistar o 18º lugar na olimpíada de matemática!"}, {"role": "assistant", "content": "Parabéns por conquistar o décimo oitavo lugar na olimpíada de matemática!"}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Houve um atraso de 38% no tempo estimado de entrega."}, {"role": "assistant", "content": "Houve um atraso de trinta e oito por cento no tempo estimado de entrega."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Você tem 9 notificações pendentes no seu portal do paciente."}, {"role": "assistant", "content": "Você tem nove notificações pendentes no seu portal do paciente."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Houve um atraso de 21,2% no tempo estimado de entrega."}, {"role": "assistant", "content": "Houve um atraso de vinte e um vírgula dois por cento no tempo estimado de entrega."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Você tem 4 notificações pendentes no seu portal do paciente."}, {"role": "assistant", "content": "Você tem quatro notificações pendentes no seu portal do paciente."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para 913 383 425."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para nove um três três oito três quatro dois cinco."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Você tem 6 notificações pendentes no seu portal do paciente."}, {"role": "assistant", "content": "Você tem seis notificações pendentes no seu portal do paciente."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Houve um atraso de 15% no tempo estimado de entrega."}, {"role": "assistant", "content": "Houve um atraso de quinze por cento no tempo estimado de entrega."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Houve um atraso de 22,7% no tempo estimado de entrega."}, {"role": "assistant", "content": "Houve um atraso de vinte e dois vírgula sete por cento no tempo estimado de entrega."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 20 - 23 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de vinte a vinte e três dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para 21 165 4491."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para dois um um seis cinco quatro quatro nove um."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Parabéns por conquistar o 1.º lugar na olimpíada de matemática!"}, {"role": "assistant", "content": "Parabéns por conquistar o primeiro lugar na olimpíada de matemática!"}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Parabéns por conquistar o 16º lugar na olimpíada de matemática!"}, {"role": "assistant", "content": "Parabéns por conquistar o décimo sexto lugar na olimpíada de matemática!"}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Parabéns por conquistar o 13.º lugar na olimpíada de matemática!"}, {"role": "assistant", "content": "Parabéns por conquistar o décimo terceiro lugar na olimpíada de matemática!"}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para 946 274 720."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para nove quatro seis dois sete quatro sete dois zero."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Houve um atraso de 62,1% no tempo estimado de entrega."}, {"role": "assistant", "content": "Houve um atraso de sessenta e dois vírgula um por cento no tempo estimado de entrega."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 14-18 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de catorze a dezoito dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Parabéns por conquistar o 31º lugar na olimpíada de matemática!"}, {"role": "assistant", "content": "Parabéns por conquistar o trigésimo primeiro lugar na olimpíada de matemática!"}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Você tem 785 notificações pendentes no seu portal do paciente."}, {"role": "assistant", "content": "Você tem setecentos e oitenta e cinco notificações pendentes no seu portal do paciente."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para 21 364 9244."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para dois um três seis quatro nove dois quatro quatro."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Você tem 901 notificações pendentes no seu portal do paciente."}, {"role": "assistant", "content": "Você tem novecentos e um notificações pendentes no seu portal do paciente."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para 21 816 6679."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para dois um oito um seis seis seis sete nove."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Houve um atraso de 91% no tempo estimado de entrega."}, {"role": "assistant", "content": "Houve um atraso de noventa e um por cento no tempo estimado de entrega."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Houve um atraso de 47% no tempo estimado de entrega."}, {"role": "assistant", "content": "Houve um atraso de quarenta e sete por cento no tempo estimado de entrega."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 27-36 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de vinte e sete a trinta e seis dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Houve um atraso de 78,2% no tempo estimado de entrega."}, {"role": "assistant", "content": "Houve um atraso de setenta e oito vírgula dois por cento no tempo estimado de entrega."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Houve um atraso de 51,1% no tempo estimado de entrega."}, {"role": "assistant", "content": "Houve um atraso de cinquenta e um vírgula um por cento no tempo estimado de entrega."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Parabéns por conquistar o 19.º lugar na olimpíada de matemática!"}, {"role": "assistant", "content": "Parabéns por conquistar o décimo nono lugar na olimpíada de matemática!"}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Houve um atraso de 87% no tempo estimado de entrega."}, {"role": "assistant", "content": "Houve um atraso de oitenta e sete por cento no tempo estimado de entrega."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 14–16 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de catorze a dezasseis dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 13-14 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de treze a catorze dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Você tem 567 notificações pendentes no seu portal do paciente."}, {"role": "assistant", "content": "Você tem quinhentos e sessenta e sete notificações pendentes no seu portal do paciente."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 9–17 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de nove a dezassete dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Houve um atraso de 22,1% no tempo estimado de entrega."}, {"role": "assistant", "content": "Houve um atraso de vinte e dois vírgula um por cento no tempo estimado de entrega."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 15-25 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de quinze a vinte e cinco dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para 21 324 3457."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para dois um três dois quatro três quatro cinco sete."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Parabéns por conquistar o 24º lugar na olimpíada de matemática!"}, {"role": "assistant", "content": "Parabéns por conquistar o vigésimo quarto lugar na olimpíada de matemática!"}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para 901 649 437."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para nove zero um seis quatro nove quatro três sete."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Você tem 116 notificações pendentes no seu portal do paciente."}, {"role": "assistant", "content": "Você tem cento e dezasseis notificações pendentes no seu portal do paciente."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 24 - 29 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de vinte e quatro a vinte e nove dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Você tem 520 notificações pendentes no seu portal do paciente."}, {"role": "assistant", "content": "Você tem quinhentos e vinte notificações pendentes no seu portal do paciente."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 14-17 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de catorze a dezassete dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para 21 746 3526."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para dois um sete quatro seis três cinco dois seis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 13–14 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de treze a catorze dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Você tem 617 notificações pendentes no seu portal do paciente."}, {"role": "assistant", "content": "Você tem seiscentos e dezassete notificações pendentes no seu portal do paciente."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Houve um atraso de 64% no tempo estimado de entrega."}, {"role": "assistant", "content": "Houve um atraso de sessenta e quatro por cento no tempo estimado de entrega."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Houve um atraso de 58% no tempo estimado de entrega."}, {"role": "assistant", "content": "Houve um atraso de cinquenta e oito por cento no tempo estimado de entrega."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para +351 912 276 437."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para mais três cinco um nove um dois dois sete seis quatro três sete."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 16 - 24 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de dezasseis a vinte e quatro dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Houve um atraso de 28% no tempo estimado de entrega."}, {"role": "assistant", "content": "Houve um atraso de vinte e oito por cento no tempo estimado de entrega."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 20-23 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de vinte a vinte e três dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Parabéns por conquistar o 16.º lugar na olimpíada de matemática!"}, {"role": "assistant", "content": "Parabéns por conquistar o décimo sexto lugar na olimpíada de matemática!"}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para 21 907 6322."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para dois um nove zero sete seis três dois dois."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Parabéns por conquistar o 26º lugar na olimpíada de matemática!"}, {"role": "assistant", "content": "Parabéns por conquistar o vigésimo sexto lugar na olimpíada de matemática!"}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para +351 956 395 255."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para mais três cinco um nove cinco seis três nove cinco dois cinco cinco."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 13–18 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de treze a dezoito dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Houve um atraso de 46% no tempo estimado de entrega."}, {"role": "assistant", "content": "Houve um atraso de quarenta e seis por cento no tempo estimado de entrega."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para 948 223 621."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para nove quatro oito dois dois três seis dois um."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Você tem 56 notificações pendentes no seu portal do paciente."}, {"role": "assistant", "content": "Você tem cinquenta e seis notificações pendentes no seu portal do paciente."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Houve um atraso de 86% no tempo estimado de entrega."}, {"role": "assistant", "content": "Houve um atraso de oitenta e seis por cento no tempo estimado de entrega."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Você tem 62 notificações pendentes no seu portal do paciente."}, {"role": "assistant", "content": "Você tem sessenta e dois notificações pendentes no seu portal do paciente."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Parabéns por conquistar o 9º lugar na olimpíada de matemática!"}, {"role": "assistant", "content": "Parabéns por conquistar o nono lugar na olimpíada de matemática!"}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 12–15 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de doze a quinze dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Você tem 742 notificações pendentes no seu portal do paciente."}, {"role": "assistant", "content": "Você tem setecentos e quarenta e dois notificações pendentes no seu portal do paciente."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 5-7 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de cinco a sete dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 19-23 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de dezanove a vinte e três dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Houve um atraso de 55% no tempo estimado de entrega."}, {"role": "assistant", "content": "Houve um atraso de cinquenta e cinco por cento no tempo estimado de entrega."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 16–18 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de dezasseis a dezoito dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para +351 969 501 053."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para mais três cinco um nove seis nove cinco zero um zero cinco três."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para +351 917 543 560."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para mais três cinco um nove um sete cinco quatro três cinco seis zero."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 5-14 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de cinco a catorze dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 28-34 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de vinte e oito a trinta e quatro dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para falar com a secretaria, ligue para 935 435 613."}, {"role": "assistant", "content": "Para falar com a secretaria, ligue para nove três cinco quatro três cinco seis um três."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Parabéns por conquistar o 28.º lugar na olimpíada de matemática!"}, {"role": "assistant", "content": "Parabéns por conquistar o vigésimo oitavo lugar na olimpíada de matemática!"}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 27-35 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de vinte e sete a trinta e cinco dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 20 - 26 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de vinte a vinte e seis dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O prazo para a entrega dos exames é de 27-33 dias úteis."}, {"role": "assistant", "content": "O prazo para a entrega dos exames é de vinte e sete a trinta e três dias úteis."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Você tem 713 notificações pendentes no seu portal do paciente."}, {"role": "assistant", "content": "Você tem setecentos e treze notificações pendentes no seu portal do paciente."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Houve um atraso de 89,1% no tempo estimado de entrega."}, {"role": "assistant", "content": "Houve um atraso de oitenta e nove vírgula um por cento no tempo estimado de entrega."}], "lang": "pt", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar pedro37@correo.es voor 27 mei 2019."}, {"role": "assistant", "content": "Stuur uw medische documenten naar pedro drie zeven at correo punt es voor zevenentwintig mei tweeduizendnegentien."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 1,1GB en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt één komma een gigabyte en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "In 1972 hebben we ons protocol voor zorg volledig vernieuwd."}, {"role": "assistant", "content": "In negentien honderd tweeënzeventig hebben we ons protocol voor zorg volledig vernieuwd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 306,8 °C en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt driehonderdzes komma acht graden Celsius en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "In 1967 hebben we ons protocol voor zorg volledig vernieuwd."}, {"role": "assistant", "content": "In negentien honderd zevenenzestig hebben we ons protocol voor zorg volledig vernieuwd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 456 graden en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt vierhonderdzesenvijftig graden Celsius en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op 17 maart bij u bezorgd."}, {"role": "assistant", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op zeventien maart bij u bezorgd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar ravi@scicom.com.my voor 1 november."}, {"role": "assistant", "content": "Stuur uw medische documenten naar ravi at scicom punt com punt my voor één november."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 1,3g en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt één komma drie gram en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op 22 december 2032 bij u bezorgd."}, {"role": "assistant", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op tweeëntwintig december tweeduizendtweeëndertig bij u bezorgd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 9,9GB en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt negen komma negen gigabyte en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar sofia@sapo.pt voor 26-12-1995."}, {"role": "assistant", "content": "Stuur uw medische documenten naar sofia at sapo punt pt voor zesentwintig december negentien honderd vijfennegentig."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 8834 te versturen naar 020 782 6963."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door acht acht drie vier te versturen naar nul twee nul zeven acht twee zes negen zes drie."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar info@yahoo.com.ph voor 28 april 2026."}, {"role": "assistant", "content": "Stuur uw medische documenten naar info at yahoo punt com punt ph voor achtentwintig april tweeduizendzesentwintig."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 6326 te versturen naar 020 779 0364."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door zes drie twee zes te versturen naar nul twee nul zeven zeven negen nul drie zes vier."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 468268 te versturen naar 020 183 5854."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door vier zes acht twee zes acht te versturen naar nul twee nul een acht drie vijf acht vijf vier."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 279GB en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt tweehonderdnegenenzeventig gigabyte en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 251066 te versturen naar +31 6 5921 7615."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door twee vijf een nul zes zes te versturen naar plus drie een zes vijf negen twee een zeven zes een vijf."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op 16-7-2027 bij u bezorgd."}, {"role": "assistant", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op zestien juli tweeduizendzevenentwintig bij u bezorgd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "In 2018 hebben we ons protocol voor zorg volledig vernieuwd."}, {"role": "assistant", "content": "In tweeduizendachttien hebben we ons protocol voor zorg volledig vernieuwd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "In 1971 hebben we ons protocol voor zorg volledig vernieuwd."}, {"role": "assistant", "content": "In negentien honderd eenenzeventig hebben we ons protocol voor zorg volledig vernieuwd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op 22-1-2029 bij u bezorgd."}, {"role": "assistant", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op tweeëntwintig januari tweeduizendnegenentwintig bij u bezorgd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op 26-12-1998 bij u bezorgd."}, {"role": "assistant", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op zesentwintig december negentien honderd achtennegentig bij u bezorgd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "In 1957 hebben we ons protocol voor zorg volledig vernieuwd."}, {"role": "assistant", "content": "In negentien honderd zevenenvijftig hebben we ons protocol voor zorg volledig vernieuwd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 9657 te versturen naar 020 058 8666."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door negen zes vijf zeven te versturen naar nul twee nul nul vijf acht acht zes zes zes."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op 22 april 1999 bij u bezorgd."}, {"role": "assistant", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op tweeëntwintig april negentien honderd negenennegentig bij u bezorgd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "In 1987 hebben we ons protocol voor zorg volledig vernieuwd."}, {"role": "assistant", "content": "In negentien honderd zevenentachtig hebben we ons protocol voor zorg volledig vernieuwd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 193g en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt honderddrieënnegentig gram en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 9g en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt negen gram en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op 24 februari 2022 bij u bezorgd."}, {"role": "assistant", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op vierentwintig februari tweeduizendtweeëntwintig bij u bezorgd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar support@example.com voor 06-06-2021."}, {"role": "assistant", "content": "Stuur uw medische documenten naar support at example punt com voor zes juni tweeduizendeenentwintig."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 6ml en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt zes milliliter en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op 14-03-1999 bij u bezorgd."}, {"role": "assistant", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op veertien maart negentien honderd negenennegentig bij u bezorgd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "In 2028 hebben we ons protocol voor zorg volledig vernieuwd."}, {"role": "assistant", "content": "In tweeduizendachtentwintig hebben we ons protocol voor zorg volledig vernieuwd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op 17 april 1999 bij u bezorgd."}, {"role": "assistant", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op zeventien april negentien honderd negenennegentig bij u bezorgd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 385 cm en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt driehonderdvijfentachtig centimeter en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 1kg en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt één kilo en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar info@gmail.co.id voor 3-8-2013."}, {"role": "assistant", "content": "Stuur uw medische documenten naar info at gmail punt co punt id voor drie augustus tweeduizenddertien."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar help@company.co voor 11 augustus."}, {"role": "assistant", "content": "Stuur uw medische documenten naar help at company punt co voor elf augustus."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "In 1993 hebben we ons protocol voor zorg volledig vernieuwd."}, {"role": "assistant", "content": "In negentien honderd drieënnegentig hebben we ons protocol voor zorg volledig vernieuwd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 450,6 g en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt vierhonderdvijftig komma zes gram en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 05224 te versturen naar 06 6553 7148."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door nul vijf twee twee vier te versturen naar nul zes zes vijf vijf drie zeven een vier acht."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar pedro19@web.de voor 15-03-2024."}, {"role": "assistant", "content": "Stuur uw medische documenten naar pedro een negen at web punt de voor vijftien maart tweeduizendvierentwintig."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 273789 te versturen naar 020 170 2550."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door twee zeven drie zeven acht negen te versturen naar nul twee nul een zeven nul twee vijf vijf nul."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 1,7l en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt één komma zeven liter en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 721351 te versturen naar +31 6 3830 2563."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door zeven twee een drie vijf een te versturen naar plus drie een zes drie acht drie nul twee vijf zes drie."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 1,5 graden en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt één komma vijf graden Celsius en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op 22 juni bij u bezorgd."}, {"role": "assistant", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op tweeëntwintig juni bij u bezorgd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 1645 te versturen naar 06 2077 0429."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door een zes vier vijf te versturen naar nul zes twee nul zeven zeven nul vier twee negen."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar help@gmail.co.id voor 04-07-2015."}, {"role": "assistant", "content": "Stuur uw medische documenten naar help at gmail punt co punt id voor vier juli tweeduizendvijftien."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op 12-02-2027 bij u bezorgd."}, {"role": "assistant", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op twaalf februari tweeduizendzevenentwintig bij u bezorgd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 1,3 km en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt één komma drie kilometer en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op 26 juni bij u bezorgd."}, {"role": "assistant", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op zesentwintig juni bij u bezorgd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 3032 te versturen naar 06 4584 5719."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door drie nul drie twee te versturen naar nul zes vier vijf acht vier vijf zeven een negen."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op 22-06-1996 bij u bezorgd."}, {"role": "assistant", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op tweeëntwintig juni negentien honderd zesennegentig bij u bezorgd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op 3 maart 2005 bij u bezorgd."}, {"role": "assistant", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op drie maart tweeduizendvijf bij u bezorgd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 459255 te versturen naar +31 6 3358 4515."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door vier vijf negen twee vijf vijf te versturen naar plus drie een zes drie drie vijf acht vier vijf een vijf."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 1,5 g en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt één komma vijf gram en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op 28-9-2032 bij u bezorgd."}, {"role": "assistant", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op achtentwintig september tweeduizendtweeëndertig bij u bezorgd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "In 2013 hebben we ons protocol voor zorg volledig vernieuwd."}, {"role": "assistant", "content": "In tweeduizenddertien hebben we ons protocol voor zorg volledig vernieuwd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "In 1994 hebben we ons protocol voor zorg volledig vernieuwd."}, {"role": "assistant", "content": "In negentien honderd vierennegentig hebben we ons protocol voor zorg volledig vernieuwd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar sofia@web.de voor 05-09-2007."}, {"role": "assistant", "content": "Stuur uw medische documenten naar sofia at web punt de voor vijf september tweeduizendzeven."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 9 m en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt negen meter en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar ravi@correo.es voor 19 november 2025."}, {"role": "assistant", "content": "Stuur uw medische documenten naar ravi at correo punt es voor negentien november tweeduizendvijfentwintig."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar support@sapo.pt voor 10-10-2018."}, {"role": "assistant", "content": "Stuur uw medische documenten naar support at sapo punt pt voor tien oktober tweeduizendachttien."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar chen88@company.co voor 28 januari 2027."}, {"role": "assistant", "content": "Stuur uw medische documenten naar chen acht acht at company punt co voor achtentwintig januari tweeduizendzevenentwintig."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 8709 te versturen naar 06 4979 0474."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door acht zeven nul negen te versturen naar nul zes vier negen zeven negen nul vier zeven vier."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "In 1974 hebben we ons protocol voor zorg volledig vernieuwd."}, {"role": "assistant", "content": "In negentien honderd vierenzeventig hebben we ons protocol voor zorg volledig vernieuwd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "In 2010 hebben we ons protocol voor zorg volledig vernieuwd."}, {"role": "assistant", "content": "In tweeduizendtien hebben we ons protocol voor zorg volledig vernieuwd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 1 cm en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt één centimeter en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar ravi19@mail.fr voor 4-11-1997."}, {"role": "assistant", "content": "Stuur uw medische documenten naar ravi een negen at mail punt fr voor vier november negentien honderd zevenennegentig."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 861926 te versturen naar 020 352 3149."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door acht zes een negen twee zes te versturen naar nul twee nul drie vijf twee drie een vier negen."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 7875 te versturen naar 06 3772 5470."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door zeven acht zeven vijf te versturen naar nul zes drie zeven zeven twee vijf vier zeven nul."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 9,2 m en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt negen komma twee meter en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 4,6km en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt vier komma zes kilometer en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 1849 te versturen naar +31 6 9086 2977."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door een acht vier negen te versturen naar plus drie een zes negen nul acht zes twee negen zeven zeven."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "In 1996 hebben we ons protocol voor zorg volledig vernieuwd."}, {"role": "assistant", "content": "In negentien honderd zesennegentig hebben we ons protocol voor zorg volledig vernieuwd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar support@gmail.com voor 10-11-2002."}, {"role": "assistant", "content": "Stuur uw medische documenten naar support at gmail punt com voor tien november tweeduizendtwee."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 8,4 ml en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt acht komma vier milliliter en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 594948 te versturen naar 06-41441772."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door vijf negen vier negen vier acht te versturen naar nul zes vier een vier vier een zeven zeven twee."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 1 kg en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt één kilo en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 7 km en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt zeven kilometer en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar ali@gmail.co.id voor 16-07-2018."}, {"role": "assistant", "content": "Stuur uw medische documenten naar ali at gmail punt co punt id voor zestien juli tweeduizendachttien."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op 01-10-2027 bij u bezorgd."}, {"role": "assistant", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op één oktober tweeduizendzevenentwintig bij u bezorgd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 6 cm en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt zes centimeter en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "In 1992 hebben we ons protocol voor zorg volledig vernieuwd."}, {"role": "assistant", "content": "In negentien honderd tweeënnegentig hebben we ons protocol voor zorg volledig vernieuwd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 205943 te versturen naar 06 9671 7007."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door twee nul vijf negen vier drie te versturen naar nul zes negen zes zeven een zeven nul nul zeven."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 318951 te versturen naar 06 1186 6588."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door drie een acht negen vijf een te versturen naar nul zes een een acht zes zes vijf acht acht."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 1 °C en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt één graad Celsius en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 35264 te versturen naar +31 6 5733 2311."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door drie vijf twee zes vier te versturen naar plus drie een zes vijf zeven drie drie twee drie een een."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar juan@sltnet.lk voor 25-05-2010."}, {"role": "assistant", "content": "Stuur uw medische documenten naar juan at sltnet punt lk voor vijfentwintig mei tweeduizendtien."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar chen88@mail.fr voor 12 mei."}, {"role": "assistant", "content": "Stuur uw medische documenten naar chen acht acht at mail punt fr voor twaalf mei."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 63767 te versturen naar 06 5235 7514."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door zes drie zeven zes zeven te versturen naar nul zes vijf twee drie vijf zeven vijf een vier."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar siti41@ziggo.nl voor 20 mei."}, {"role": "assistant", "content": "Stuur uw medische documenten naar siti vier een at ziggo punt nl voor twintig mei."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 167 g en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt honderdzevenenzestig gram en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 6 GB en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt zes gigabyte en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 6081 te versturen naar +31 6 7197 7568."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door zes nul acht een te versturen naar plus drie een zes zeven een negen zeven zeven vijf zes acht."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 4844 te versturen naar 020 569 9186."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door vier acht vier vier te versturen naar nul twee nul vijf zes negen negen een acht zes."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 2 GB en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt twee gigabyte en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 8m en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt acht meter en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar jan.devries@scicom.com.my voor 17-7-2019."}, {"role": "assistant", "content": "Stuur uw medische documenten naar jan punt devries at scicom punt com punt my voor zeventien juli tweeduizendnegentien."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "In 2022 hebben we ons protocol voor zorg volledig vernieuwd."}, {"role": "assistant", "content": "In tweeduizendtweeëntwintig hebben we ons protocol voor zorg volledig vernieuwd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar luca@sltnet.lk voor 26 april."}, {"role": "assistant", "content": "Stuur uw medische documenten naar luca at sltnet punt lk voor zesentwintig april."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op 12-9-2032 bij u bezorgd."}, {"role": "assistant", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op twaalf september tweeduizendtweeëndertig bij u bezorgd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "In 1999 hebben we ons protocol voor zorg volledig vernieuwd."}, {"role": "assistant", "content": "In negentien honderd negenennegentig hebben we ons protocol voor zorg volledig vernieuwd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar ahmad.b@gmail.co.id voor 2 januari."}, {"role": "assistant", "content": "Stuur uw medische documenten naar ahmad punt b at gmail punt co punt id voor twee januari."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar maria59@yahoo.com.ph voor 13-1-2007."}, {"role": "assistant", "content": "Stuur uw medische documenten naar maria vijf negen at yahoo punt com punt ph voor dertien januari tweeduizendzeven."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 867726 te versturen naar 06 1338 0533."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door acht zes zeven zeven twee zes te versturen naar nul zes een drie drie acht nul vijf drie drie."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar jan.devries@mail.fr voor 11 juni."}, {"role": "assistant", "content": "Stuur uw medische documenten naar jan punt devries at mail punt fr voor elf juni."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "In 2007 hebben we ons protocol voor zorg volledig vernieuwd."}, {"role": "assistant", "content": "In tweeduizendzeven hebben we ons protocol voor zorg volledig vernieuwd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op 24 augustus bij u bezorgd."}, {"role": "assistant", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op vierentwintig augustus bij u bezorgd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar support@example.com voor 6-3-2027."}, {"role": "assistant", "content": "Stuur uw medische documenten naar support at example punt com voor zes maart tweeduizendzevenentwintig."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 8,2 L en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt acht komma twee liter en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op 14 juni 1999 bij u bezorgd."}, {"role": "assistant", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op veertien juni negentien honderd negenennegentig bij u bezorgd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 1114 te versturen naar 020 689 4942."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door een een een vier te versturen naar nul twee nul zes acht negen vier negen vier twee."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar maria@onet.pl voor 13 april 2028."}, {"role": "assistant", "content": "Stuur uw medische documenten naar maria at onet punt pl voor dertien april tweeduizendachtentwintig."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar help@company.co voor 15 augustus."}, {"role": "assistant", "content": "Stuur uw medische documenten naar help at company punt co voor vijftien augustus."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 8273 te versturen naar 06-95362790."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door acht twee zeven drie te versturen naar nul zes negen vijf drie zes twee zeven negen nul."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op 28 oktober 2013 bij u bezorgd."}, {"role": "assistant", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op achtentwintig oktober tweeduizenddertien bij u bezorgd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "In 2027 hebben we ons protocol voor zorg volledig vernieuwd."}, {"role": "assistant", "content": "In tweeduizendzevenentwintig hebben we ons protocol voor zorg volledig vernieuwd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar juan@bank.my voor 8 juli."}, {"role": "assistant", "content": "Stuur uw medische documenten naar juan at bank punt my voor acht juli."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar omar27@onet.pl voor 26-7-1995."}, {"role": "assistant", "content": "Stuur uw medische documenten naar omar twee zeven at onet punt pl voor zesentwintig juli negentien honderd vijfennegentig."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op 4-1-2031 bij u bezorgd."}, {"role": "assistant", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op vier januari tweeduizendeenendertig bij u bezorgd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "In 2034 hebben we ons protocol voor zorg volledig vernieuwd."}, {"role": "assistant", "content": "In tweeduizendvierendertig hebben we ons protocol voor zorg volledig vernieuwd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "In 1995 hebben we ons protocol voor zorg volledig vernieuwd."}, {"role": "assistant", "content": "In negentien honderd vijfennegentig hebben we ons protocol voor zorg volledig vernieuwd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op 23 september 2006 bij u bezorgd."}, {"role": "assistant", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op drieëntwintig september tweeduizendzes bij u bezorgd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 752120 te versturen naar +31 6 0542 1586."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door zeven vijf twee een twee nul te versturen naar plus drie een zes nul vijf vier twee een vijf acht zes."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "In 2032 hebben we ons protocol voor zorg volledig vernieuwd."}, {"role": "assistant", "content": "In tweeduizendtweeëndertig hebben we ons protocol voor zorg volledig vernieuwd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 371325 te versturen naar 06-81701354."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door drie zeven een drie twee vijf te versturen naar nul zes acht een zeven nul een drie vijf vier."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "In 1955 hebben we ons protocol voor zorg volledig vernieuwd."}, {"role": "assistant", "content": "In negentien honderd vijfenvijftig hebben we ons protocol voor zorg volledig vernieuwd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 80303 te versturen naar 06-30777411."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door acht nul drie nul drie te versturen naar nul zes drie nul zeven zeven zeven vier een een."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 156cm en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt honderdzesenvijftig centimeter en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 1cm en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt één centimeter en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "In 1954 hebben we ons protocol voor zorg volledig vernieuwd."}, {"role": "assistant", "content": "In negentien honderd vierenvijftig hebben we ons protocol voor zorg volledig vernieuwd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op 21 augustus bij u bezorgd."}, {"role": "assistant", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op eenentwintig augustus bij u bezorgd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 3°C en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt drie graden Celsius en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 1 GB en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt één gigabyte en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "In 1962 hebben we ons protocol voor zorg volledig vernieuwd."}, {"role": "assistant", "content": "In negentien honderd tweeënzestig hebben we ons protocol voor zorg volledig vernieuwd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 1,5 m en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt één komma vijf meter en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw pakket weegt 1,6 GB en is nu onderweg naar het sorteercentrum."}, {"role": "assistant", "content": "Uw pakket weegt één komma zes gigabyte en is nu onderweg naar het sorteercentrum."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar ali63@web.de voor 27-08-1997."}, {"role": "assistant", "content": "Stuur uw medische documenten naar ali zes drie at web punt de voor zevenentwintig augustus negentien honderd zevenennegentig."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 7911 te versturen naar 06-89293401."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door zeven negen een een te versturen naar nul zes acht negen twee negen drie vier nul een."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op 9 mei 2003 bij u bezorgd."}, {"role": "assistant", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op negen mei tweeduizenddrie bij u bezorgd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Stuur uw medische documenten naar dilani@yahoo.com.ph voor 3-1-2032."}, {"role": "assistant", "content": "Stuur uw medische documenten naar dilani at yahoo punt com punt ph voor drie januari tweeduizendtweeëndertig."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op 15-02-2020 bij u bezorgd."}, {"role": "assistant", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op vijftien februari tweeduizendtwintig bij u bezorgd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 865601 te versturen naar +31 6 7099 7516."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door acht zes vijf zes nul een te versturen naar plus drie een zes zeven nul negen negen zeven vijf een zes."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op 5 augustus bij u bezorgd."}, {"role": "assistant", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op vijf augustus bij u bezorgd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 168339 te versturen naar +31 6 4376 6829."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door een zes acht drie drie negen te versturen naar plus drie een zes vier drie zeven zes zes acht twee negen."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U kunt uw nieuwe abonnement activeren door 2375 te versturen naar 06 5980 3244."}, {"role": "assistant", "content": "U kunt uw nieuwe abonnement activeren door twee drie zeven vijf te versturen naar nul zes vijf negen acht nul drie twee vier vier."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op 12-6-2023 bij u bezorgd."}, {"role": "assistant", "content": "Uw nieuwe simkaart wordt verzonden en is naar verwachting op twaalf juni tweeduizenddrieëntwintig bij u bezorgd."}], "lang": "nl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę PLN 70 534 jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę siedemdziesiąt tysięcy pięćset trzydzieści cztery złote jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja oferta wygasa za 4 dni."}, {"role": "assistant", "content": "Twoja oferta wygasa za cztery dni."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Więcej informacji znajdziesz na stronie https://scicom.com.my/faq."}, {"role": "assistant", "content": "Więcej informacji znajdziesz na stronie scicom kropka com kropka my ukośnik faq."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Więcej informacji znajdziesz na stronie https://outlook.sa/help."}, {"role": "assistant", "content": "Więcej informacji znajdziesz na stronie outlook kropka sa ukośnik help."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę 8 418 zł jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę osiem tysięcy czterysta osiemnaście złotych jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja oferta wygasa za 5 dni."}, {"role": "assistant", "content": "Twoja oferta wygasa za pięć dni."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja oferta wygasa za 1 dni."}, {"role": "assistant", "content": "Twoja oferta wygasa za jeden dni."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twój numer zgłoszenia to INV0472, prosimy podać go przy kontakcie."}, {"role": "assistant", "content": "Twój numer zgłoszenia to INV zero cztery siedem dwa, prosimy podać go przy kontakcie."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę 942 zł jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę dziewięćset czterdzieści dwa złote jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twój numer zgłoszenia to ABC1234, prosimy podać go przy kontakcie."}, {"role": "assistant", "content": "Twój numer zgłoszenia to ABC jeden dwa trzy cztery, prosimy podać go przy kontakcie."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twój numer zgłoszenia to WXY 1234, prosimy podać go przy kontakcie."}, {"role": "assistant", "content": "Twój numer zgłoszenia to WXY jeden dwa trzy cztery, prosimy podać go przy kontakcie."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę PLN 847,12 jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę osiemset czterdzieści siedem złotych dwanaście groszy jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer +48 565 356 363."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer plus cztery osiem pięć sześć pięć trzy pięć sześć trzy sześć trzy."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 672 184 362."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer sześć siedem dwa jeden osiem cztery trzy sześć dwa."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Więcej informacji znajdziesz na stronie https://sltnet.lk."}, {"role": "assistant", "content": "Więcej informacji znajdziesz na stronie sltnet kropka lk."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja oferta wygasa za 40 dni."}, {"role": "assistant", "content": "Twoja oferta wygasa za czterdzieści dni."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer +48 556 830 695."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer plus cztery osiem pięć pięć sześć osiem trzy zero sześć dziewięć pięć."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 510 115 853."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer pięć jeden zero jeden jeden pięć osiem pięć trzy."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 637 107 401."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer sześć trzy siedem jeden zero siedem cztery zero jeden."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Więcej informacji znajdziesz na stronie www.example.com/login."}, {"role": "assistant", "content": "Więcej informacji znajdziesz na stronie w w w kropka example kropka com ukośnik login."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Więcej informacji znajdziesz na stronie https://www.gmail.com/support."}, {"role": "assistant", "content": "Więcej informacji znajdziesz na stronie w w w kropka gmail kropka com ukośnik support."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę 45,98 zł jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę czterdzieści pięć złotych dziewięćdziesiąt osiem groszy jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twój numer zgłoszenia to QZ8501, prosimy podać go przy kontakcie."}, {"role": "assistant", "content": "Twój numer zgłoszenia to QZ osiem pięć zero jeden, prosimy podać go przy kontakcie."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę 443 zł jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę czterysta czterdzieści trzy złote jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 22 709 15 77."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer dwa dwa siedem zero dziewięć jeden pięć siedem siedem."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę PLN 85,01 jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę osiemdziesiąt pięć złotych jeden grosz jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja oferta wygasa za 837 dni."}, {"role": "assistant", "content": "Twoja oferta wygasa za osiemset trzydzieści siedem dni."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Więcej informacji znajdziesz na stronie https://outlook.sa."}, {"role": "assistant", "content": "Więcej informacji znajdziesz na stronie outlook kropka sa."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twój numer zgłoszenia to B2-14, prosimy podać go przy kontakcie."}, {"role": "assistant", "content": "Twój numer zgłoszenia to B dwa jeden cztery, prosimy podać go przy kontakcie."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 574 211 787."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer pięć siedem cztery dwa jeden jeden siedem osiem siedem."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę PLN 526,72 jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę pięćset dwadzieścia sześć złotych siedemdziesiąt dwa grosze jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja oferta wygasa za 6 dni."}, {"role": "assistant", "content": "Twoja oferta wygasa za sześć dni."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę 158,50 zł jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę sto pięćdziesiąt osiem złotych pięćdziesiąt groszy jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 675 372 989."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer sześć siedem pięć trzy siedem dwa dziewięć osiem dziewięć."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja oferta wygasa za 24 dni."}, {"role": "assistant", "content": "Twoja oferta wygasa za dwadzieścia cztery dni."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer +48 549 894 462."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer plus cztery osiem pięć cztery dziewięć osiem dziewięć cztery cztery sześć dwa."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 647 090 555."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer sześć cztery siedem zero dziewięć zero pięć pięć pięć."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 609 110 666."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer sześć zero dziewięć jeden jeden zero sześć sześć sześć."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 584 225 537."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer pięć osiem cztery dwa dwa pięć pięć trzy siedem."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę 95 zł jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę dziewięćdziesiąt pięć złotych jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę 53 691,90 zł jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę pięćdziesiąt trzy tysiące sześćset dziewięćdziesiąt jeden złotych dziewięćdziesiąt groszy jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Więcej informacji znajdziesz na stronie www.company.co/login."}, {"role": "assistant", "content": "Więcej informacji znajdziesz na stronie w w w kropka company kropka co ukośnik login."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja oferta wygasa za 43 dni."}, {"role": "assistant", "content": "Twoja oferta wygasa za czterdzieści trzy dni."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja oferta wygasa za 9 dni."}, {"role": "assistant", "content": "Twoja oferta wygasa za dziewięć dni."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę PLN 129,90 jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę sto dwadzieścia dziewięć złotych dziewięćdziesiąt groszy jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja oferta wygasa za 2 dni."}, {"role": "assistant", "content": "Twoja oferta wygasa za dwa dni."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Więcej informacji znajdziesz na stronie www.posta.it."}, {"role": "assistant", "content": "Więcej informacji znajdziesz na stronie w w w kropka posta kropka it."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer +48 532 846 204."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer plus cztery osiem pięć trzy dwa osiem cztery sześć dwa zero cztery."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Więcej informacji znajdziesz na stronie www.correo.es."}, {"role": "assistant", "content": "Więcej informacji znajdziesz na stronie w w w kropka correo kropka es."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer +48 504 874 757."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer plus cztery osiem pięć zero cztery osiem siedem cztery siedem pięć siedem."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja oferta wygasa za 3 dni."}, {"role": "assistant", "content": "Twoja oferta wygasa za trzy dni."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 653 105 214."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer sześć pięć trzy jeden zero pięć dwa jeden cztery."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twój numer zgłoszenia to X-12340567, prosimy podać go przy kontakcie."}, {"role": "assistant", "content": "Twój numer zgłoszenia to X jeden dwa trzy cztery zero pięć sześć siedem, prosimy podać go przy kontakcie."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja oferta wygasa za 58 dni."}, {"role": "assistant", "content": "Twoja oferta wygasa za pięćdziesiąt osiem dni."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Więcej informacji znajdziesz na stronie https://web.de/help."}, {"role": "assistant", "content": "Więcej informacji znajdziesz na stronie web kropka de ukośnik help."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 22 928 81 53."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer dwa dwa dziewięć dwa osiem osiem jeden pięć trzy."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę PLN 162,90 jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę sto sześćdziesiąt dwa złote dziewięćdziesiąt groszy jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 525 824 073."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer pięć dwa pięć osiem dwa cztery zero siedem trzy."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja oferta wygasa za 41 dni."}, {"role": "assistant", "content": "Twoja oferta wygasa za czterdzieści jeden dni."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja oferta wygasa za 59 dni."}, {"role": "assistant", "content": "Twoja oferta wygasa za pięćdziesiąt dziewięć dni."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 22 676 38 52."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer dwa dwa sześć siedem sześć trzy osiem pięć dwa."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja oferta wygasa za 129 dni."}, {"role": "assistant", "content": "Twoja oferta wygasa za sto dwadzieścia dziewięć dni."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 653 982 385."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer sześć pięć trzy dziewięć osiem dwa trzy osiem pięć."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Więcej informacji znajdziesz na stronie https://gmail.com."}, {"role": "assistant", "content": "Więcej informacji znajdziesz na stronie gmail kropka com."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Więcej informacji znajdziesz na stronie https://www.outlook.sa/support."}, {"role": "assistant", "content": "Więcej informacji znajdziesz na stronie w w w kropka outlook kropka sa ukośnik support."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 605 220 160."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer sześć zero pięć dwa dwa zero jeden sześć zero."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 660 982 896."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer sześć sześć zero dziewięć osiem dwa osiem dziewięć sześć."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja oferta wygasa za 181 dni."}, {"role": "assistant", "content": "Twoja oferta wygasa za sto osiemdziesiąt jeden dni."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę 942,90 zł jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę dziewięćset czterdzieści dwa złote dziewięćdziesiąt groszy jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja oferta wygasa za 971 dni."}, {"role": "assistant", "content": "Twoja oferta wygasa za dziewięćset siedemdziesiąt jeden dni."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Więcej informacji znajdziesz na stronie https://www.gmail.co.id/support."}, {"role": "assistant", "content": "Więcej informacji znajdziesz na stronie w w w kropka gmail kropka co kropka id ukośnik support."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja oferta wygasa za 148 dni."}, {"role": "assistant", "content": "Twoja oferta wygasa za sto czterdzieści osiem dni."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer +48 567 970 796."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer plus cztery osiem pięć sześć siedem dziewięć siedem zero siedem dziewięć sześć."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Więcej informacji znajdziesz na stronie https://ziggo.nl/help."}, {"role": "assistant", "content": "Więcej informacji znajdziesz na stronie ziggo kropka nl ukośnik help."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 22 228 92 24."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer dwa dwa dwa dwa osiem dziewięć dwa dwa cztery."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę 83 081,50 zł jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę osiemdziesiąt trzy tysiące osiemdziesiąt jeden złotych pięćdziesiąt groszy jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twój numer zgłoszenia to MH370, prosimy podać go przy kontakcie."}, {"role": "assistant", "content": "Twój numer zgłoszenia to MH trzy siedem zero, prosimy podać go przy kontakcie."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 22 200 14 05."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer dwa dwa dwa zero zero jeden cztery zero pięć."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Więcej informacji znajdziesz na stronie https://example.com."}, {"role": "assistant", "content": "Więcej informacji znajdziesz na stronie example kropka com."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę 840,90 zł jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę osiemset czterdzieści złotych dziewięćdziesiąt groszy jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer +48 508 894 813."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer plus cztery osiem pięć zero osiem osiem dziewięć cztery osiem jeden trzy."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 616 327 683."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer sześć jeden sześć trzy dwa siedem sześć osiem trzy."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Więcej informacji znajdziesz na stronie www.sapo.pt/login."}, {"role": "assistant", "content": "Więcej informacji znajdziesz na stronie w w w kropka sapo kropka pt ukośnik login."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę PLN 27,50 jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę dwadzieścia siedem złotych pięćdziesiąt groszy jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę PLN 70 232 jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę siedemdziesiąt tysięcy dwieście trzydzieści dwa złote jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja oferta wygasa za 69 dni."}, {"role": "assistant", "content": "Twoja oferta wygasa za sześćdziesiąt dziewięć dni."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę 72zł jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę siedemdziesiąt dwa złote jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twój numer zgłoszenia to A12, prosimy podać go przy kontakcie."}, {"role": "assistant", "content": "Twój numer zgłoszenia to A jeden dwa, prosimy podać go przy kontakcie."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Więcej informacji znajdziesz na stronie https://gmail.com/faq."}, {"role": "assistant", "content": "Więcej informacji znajdziesz na stronie gmail kropka com ukośnik faq."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer +48 543 329 582."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer plus cztery osiem pięć cztery trzy trzy dwa dziewięć pięć osiem dwa."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja oferta wygasa za 693 dni."}, {"role": "assistant", "content": "Twoja oferta wygasa za sześćset dziewięćdziesiąt trzy dni."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 654 392 899."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer sześć pięć cztery trzy dziewięć dwa osiem dziewięć dziewięć."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę PLN 289,90 jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę dwieście osiemdziesiąt dziewięć złotych dziewięćdziesiąt groszy jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 569 125 669."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer pięć sześć dziewięć jeden dwa pięć sześć sześć dziewięć."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 682 804 200."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer sześć osiem dwa osiem zero cztery dwa zero zero."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę 23 391,90 zł jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę dwadzieścia trzy tysiące trzysta dziewięćdziesiąt jeden złotych dziewięćdziesiąt groszy jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę 32 zł jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę trzydzieści dwa złote jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Więcej informacji znajdziesz na stronie https://www.yahoo.com.ph/support."}, {"role": "assistant", "content": "Więcej informacji znajdziesz na stronie w w w kropka yahoo kropka com kropka ph ukośnik support."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 22 112 69 73."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer dwa dwa jeden jeden dwa sześć dziewięć siedem trzy."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę 40 941,90 zł jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę czterdzieści tysięcy dziewięćset czterdzieści jeden złotych dziewięćdziesiąt groszy jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Więcej informacji znajdziesz na stronie www.sltnet.lk."}, {"role": "assistant", "content": "Więcej informacji znajdziesz na stronie w w w kropka sltnet kropka lk."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 22 839 05 04."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer dwa dwa osiem trzy dziewięć zero pięć zero cztery."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę 659,90zł jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę sześćset pięćdziesiąt dziewięć złotych dziewięćdziesiąt groszy jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twój numer zgłoszenia to AK6120, prosimy podać go przy kontakcie."}, {"role": "assistant", "content": "Twój numer zgłoszenia to AK sześć jeden dwa zero, prosimy podać go przy kontakcie."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 22 291 96 38."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer dwa dwa dwa dziewięć jeden dziewięć sześć trzy osiem."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę PLN 629 jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę sześćset dwadzieścia dziewięć złotych jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę 52,90 zł jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę pięćdziesiąt dwa złote dziewięćdziesiąt groszy jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 525 926 229."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer pięć dwa pięć dziewięć dwa sześć dwa dwa dziewięć."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 629 601 316."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer sześć dwa dziewięć sześć zero jeden trzy jeden sześć."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę 357,90 zł jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę trzysta pięćdziesiąt siedem złotych dziewięćdziesiąt groszy jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Więcej informacji znajdziesz na stronie https://www.onet.pl/support."}, {"role": "assistant", "content": "Więcej informacji znajdziesz na stronie w w w kropka onet kropka pl ukośnik support."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Więcej informacji znajdziesz na stronie https://company.co/help."}, {"role": "assistant", "content": "Więcej informacji znajdziesz na stronie company kropka co ukośnik help."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 618 808 069."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer sześć jeden osiem osiem zero osiem zero sześć dziewięć."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę 86 zł jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę osiemdziesiąt sześć złotych jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twój numer zgłoszenia to PO 55219, prosimy podać go przy kontakcie."}, {"role": "assistant", "content": "Twój numer zgłoszenia to PO pięć pięć dwa jeden dziewięć, prosimy podać go przy kontakcie."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 22 542 41 89."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer dwa dwa pięć cztery dwa cztery jeden osiem dziewięć."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Więcej informacji znajdziesz na stronie https://yahoo.com.ph/faq."}, {"role": "assistant", "content": "Więcej informacji znajdziesz na stronie yahoo kropka com kropka ph ukośnik faq."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę 147,94 zł jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę sto czterdzieści siedem złotych dziewięćdziesiąt cztery grosze jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę PLN 66 088 jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę sześćdziesiąt sześć tysięcy osiemdziesiąt osiem złotych jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "W celu aktywacji usługi zadzwoń pod numer 22 293 89 07."}, {"role": "assistant", "content": "W celu aktywacji usługi zadzwoń pod numer dwa dwa dwa dziewięć trzy osiem dziewięć zero siedem."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura na kwotę PLN 724,50 jest już dostępna w panelu klienta."}, {"role": "assistant", "content": "Twoja faktura na kwotę siedemset dwadzieścia cztery złote pięćdziesiąt groszy jest już dostępna w panelu klienta."}], "lang": "pl", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 19.7% untuk spending lebih RM326.50 sebulan."}, {"role": "assistant", "content": "Cash back sembilan belas perpuluhan tujuh peratus untuk spending lebih tiga ratus dua puluh enam ringgit lima puluh sen sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 54% untuk spending lebih RM 63.23 sebulan."}, {"role": "assistant", "content": "Cash back lima puluh empat peratus untuk spending lebih enam puluh tiga ringgit dua puluh tiga sen sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 11:20 sampai 2:45, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul sebelas dua puluh sampai dua empat puluh lima, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 18:45 sampai 2:05, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul lapan belas empat puluh lima sampai dua kosong lima, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 3:00 petang sampai 5.00 petang, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul tiga petang sampai lima petang, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 72.8% untuk spending lebih RM1.90 sebulan."}, {"role": "assistant", "content": "Cash back tujuh puluh dua perpuluhan lapan peratus untuk spending lebih satu ringgit sembilan puluh sen sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 5% untuk spending lebih RM75,168 sebulan."}, {"role": "assistant", "content": "Cash back lima peratus untuk spending lebih tujuh puluh lima ribu seratus enam puluh lapan ringgit sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 24.9% untuk spending lebih RM55,564.50 sebulan."}, {"role": "assistant", "content": "Cash back dua puluh empat perpuluhan sembilan peratus untuk spending lebih lima puluh lima ribu lima ratus enam puluh empat ringgit lima puluh sen sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 90% untuk spending lebih RM 38 sebulan."}, {"role": "assistant", "content": "Cash back sembilan puluh peratus untuk spending lebih tiga puluh lapan ringgit sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 9:20 pagi sampai 3.00 malam, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul sembilan dua puluh pagi sampai tiga malam, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 60.7% untuk spending lebih RM96,047 sebulan."}, {"role": "assistant", "content": "Cash back enam puluh perpuluhan tujuh peratus untuk spending lebih sembilan puluh enam ribu empat puluh tujuh ringgit sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 18% untuk spending lebih RM79,378 sebulan."}, {"role": "assistant", "content": "Cash back lapan belas peratus untuk spending lebih tujuh puluh sembilan ribu tiga ratus tujuh puluh lapan ringgit sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 12.00 petang sampai 4:20 malam, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul dua belas petang sampai empat dua puluh malam, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 11 pagi sampai 4:20 malam, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul sebelas pagi sampai empat dua puluh malam, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 9:05 malam sampai 6.45 pagi, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul sembilan kosong lima malam sampai enam empat puluh lima pagi, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 66% untuk spending lebih RM643 sebulan."}, {"role": "assistant", "content": "Cash back enam puluh enam peratus untuk spending lebih enam ratus empat puluh tiga ringgit sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 1.20 malam sampai 2:30 malam, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul satu dua puluh malam sampai dua tiga puluh malam, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 51.4% untuk spending lebih RM 82,105.90 sebulan."}, {"role": "assistant", "content": "Cash back lima puluh satu perpuluhan empat peratus untuk spending lebih lapan puluh dua ribu seratus lima ringgit sembilan puluh sen sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 16.7% untuk spending lebih RM32.50 sebulan."}, {"role": "assistant", "content": "Cash back enam belas perpuluhan tujuh peratus untuk spending lebih tiga puluh dua ringgit lima puluh sen sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 12 petang sampai 11 malam, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul dua belas petang sampai sebelas malam, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 39% untuk spending lebih RM68 sebulan."}, {"role": "assistant", "content": "Cash back tiga puluh sembilan peratus untuk spending lebih enam puluh lapan ringgit sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 2.20 petang sampai 10 pagi, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul dua dua puluh petang sampai sepuluh pagi, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 6% untuk spending lebih RM 1.79 sebulan."}, {"role": "assistant", "content": "Cash back enam peratus untuk spending lebih satu ringgit tujuh puluh sembilan sen sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 13:50 sampai 3:30, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul tiga belas lima puluh sampai tiga tiga puluh, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 13% untuk spending lebih RM 97.50 sebulan."}, {"role": "assistant", "content": "Cash back tiga belas peratus untuk spending lebih sembilan puluh tujuh ringgit lima puluh sen sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 50% untuk spending lebih RM23 sebulan."}, {"role": "assistant", "content": "Cash back lima puluh peratus untuk spending lebih dua puluh tiga ringgit sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 7.05 malam sampai 8.45 pagi, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul tujuh kosong lima malam sampai lapan empat puluh lima pagi, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 6% untuk spending lebih RM 391 sebulan."}, {"role": "assistant", "content": "Cash back enam peratus untuk spending lebih tiga ratus sembilan puluh satu ringgit sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 58.4% untuk spending lebih RM9.90 sebulan."}, {"role": "assistant", "content": "Cash back lima puluh lapan perpuluhan empat peratus untuk spending lebih sembilan ringgit sembilan puluh sen sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 76% untuk spending lebih RM 780.90 sebulan."}, {"role": "assistant", "content": "Cash back tujuh puluh enam peratus untuk spending lebih tujuh ratus lapan puluh ringgit sembilan puluh sen sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 57% untuk spending lebih RM622.90 sebulan."}, {"role": "assistant", "content": "Cash back lima puluh tujuh peratus untuk spending lebih enam ratus dua puluh dua ringgit sembilan puluh sen sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 11:50 malam sampai 16:00, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul sebelas lima puluh malam sampai enam belas, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 2.10 petang sampai 4:00 petang, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul dua sepuluh petang sampai empat petang, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 9:50 pagi sampai 9 malam, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul sembilan lima puluh pagi sampai sembilan malam, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 10.00 malam sampai 10.05 malam, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul sepuluh malam sampai sepuluh kosong lima malam, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 92.3% untuk spending lebih RM 55,847.90 sebulan."}, {"role": "assistant", "content": "Cash back sembilan puluh dua perpuluhan tiga peratus untuk spending lebih lima puluh lima ribu lapan ratus empat puluh tujuh ringgit sembilan puluh sen sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 58% untuk spending lebih RM 68,819 sebulan."}, {"role": "assistant", "content": "Cash back lima puluh lapan peratus untuk spending lebih enam puluh lapan ribu lapan ratus sembilan belas ringgit sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 6.05 pagi sampai 5.15 petang, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul enam kosong lima pagi sampai lima lima belas petang, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 2:50 petang sampai 1 petang, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul dua lima puluh petang sampai satu petang, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 93% untuk spending lebih RM85 sebulan."}, {"role": "assistant", "content": "Cash back sembilan puluh tiga peratus untuk spending lebih lapan puluh lima ringgit sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 65.8% untuk spending lebih RM14,968 sebulan."}, {"role": "assistant", "content": "Cash back enam puluh lima perpuluhan lapan peratus untuk spending lebih empat belas ribu sembilan ratus enam puluh lapan ringgit sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 16:20 sampai 9:50, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul enam belas dua puluh sampai sembilan lima puluh, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 67% untuk spending lebih RM 2.90 sebulan."}, {"role": "assistant", "content": "Cash back enam puluh tujuh peratus untuk spending lebih dua ringgit sembilan puluh sen sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 11:05 malam sampai 12:00, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul sebelas kosong lima malam sampai dua belas, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 55% untuk spending lebih RM647 sebulan."}, {"role": "assistant", "content": "Cash back lima puluh lima peratus untuk spending lebih enam ratus empat puluh tujuh ringgit sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 12 malam sampai 9:30 malam, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul dua belas malam sampai sembilan tiga puluh malam, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 41.3% untuk spending lebih RM36.50 sebulan."}, {"role": "assistant", "content": "Cash back empat puluh satu perpuluhan tiga peratus untuk spending lebih tiga puluh enam ringgit lima puluh sen sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 4 malam sampai 5:00, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul empat malam sampai lima, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 10.15 malam sampai 2:10 malam, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul sepuluh lima belas malam sampai dua sepuluh malam, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 99% untuk spending lebih RM 3,248.50 sebulan."}, {"role": "assistant", "content": "Cash back sembilan puluh sembilan peratus untuk spending lebih tiga ribu dua ratus empat puluh lapan ringgit lima puluh sen sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 36% untuk spending lebih RM 74,124 sebulan."}, {"role": "assistant", "content": "Cash back tiga puluh enam peratus untuk spending lebih tujuh puluh empat ribu seratus dua puluh empat ringgit sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 12 malam sampai 3.15 petang, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul dua belas malam sampai tiga lima belas petang, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 2.20 malam sampai 23:15, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul dua dua puluh malam sampai dua puluh tiga lima belas, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 7:10 malam sampai 10:45, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul tujuh sepuluh malam sampai sepuluh empat puluh lima, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 67.9% untuk spending lebih RM 48,645 sebulan."}, {"role": "assistant", "content": "Cash back enam puluh tujuh perpuluhan sembilan peratus untuk spending lebih empat puluh lapan ribu enam ratus empat puluh lima ringgit sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 0:20 sampai 7 pagi, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul sifar dua puluh sampai tujuh pagi, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 51.5% untuk spending lebih RM32.69 sebulan."}, {"role": "assistant", "content": "Cash back lima puluh satu perpuluhan lima peratus untuk spending lebih tiga puluh dua ringgit enam puluh sembilan sen sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 9:30 pagi sampai 7:50 pagi, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul sembilan tiga puluh pagi sampai tujuh lima puluh pagi, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 36.5% untuk spending lebih RM 733.90 sebulan."}, {"role": "assistant", "content": "Cash back tiga puluh enam perpuluhan lima peratus untuk spending lebih tujuh ratus tiga puluh tiga ringgit sembilan puluh sen sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 59% untuk spending lebih RM84 sebulan."}, {"role": "assistant", "content": "Cash back lima puluh sembilan peratus untuk spending lebih lapan puluh empat ringgit sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 17% untuk spending lebih RM 77.47 sebulan."}, {"role": "assistant", "content": "Cash back tujuh belas peratus untuk spending lebih tujuh puluh tujuh ringgit empat puluh tujuh sen sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 3.10 malam sampai 2.20 petang, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul tiga sepuluh malam sampai dua dua puluh petang, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 3:20 petang sampai 4 petang, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul tiga dua puluh petang sampai empat petang, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 8.10 malam sampai 16:20, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul lapan sepuluh malam sampai enam belas dua puluh, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 68.3% untuk spending lebih RM 58,500 sebulan."}, {"role": "assistant", "content": "Cash back enam puluh lapan perpuluhan tiga peratus untuk spending lebih lima puluh lapan ribu lima ratus ringgit sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 13:40 sampai 11:20 malam, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul tiga belas empat puluh sampai sebelas dua puluh malam, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 45% untuk spending lebih RM74.07 sebulan."}, {"role": "assistant", "content": "Cash back empat puluh lima peratus untuk spending lebih tujuh puluh empat ringgit tujuh sen sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 12.45 petang sampai 4 petang, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul dua belas empat puluh lima petang sampai empat petang, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 39% untuk spending lebih RM53 sebulan."}, {"role": "assistant", "content": "Cash back tiga puluh sembilan peratus untuk spending lebih lima puluh tiga ringgit sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 76% untuk spending lebih RM 931.90 sebulan."}, {"role": "assistant", "content": "Cash back tujuh puluh enam peratus untuk spending lebih sembilan ratus tiga puluh satu ringgit sembilan puluh sen sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 5.15 petang sampai 4.15 petang, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul lima lima belas petang sampai empat lima belas petang, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 16:15 sampai 21:45, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul enam belas lima belas sampai dua puluh satu empat puluh lima, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 10.10 malam sampai 12:00 petang, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul sepuluh sepuluh malam sampai dua belas petang, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 12 petang sampai 4:30, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul dua belas petang sampai empat tiga puluh, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 1 malam sampai 9.00 malam, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul satu malam sampai sembilan malam, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 22.5% untuk spending lebih RM 11 sebulan."}, {"role": "assistant", "content": "Cash back dua puluh dua perpuluhan lima peratus untuk spending lebih sebelas ringgit sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 11:45 sampai 11 malam, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul sebelas empat puluh lima sampai sebelas malam, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 9.45 malam sampai 0:15, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul sembilan empat puluh lima malam sampai sifar lima belas, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 44.7% untuk spending lebih RM462 sebulan."}, {"role": "assistant", "content": "Cash back empat puluh empat perpuluhan tujuh peratus untuk spending lebih empat ratus enam puluh dua ringgit sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 19% untuk spending lebih RM 27,527 sebulan."}, {"role": "assistant", "content": "Cash back sembilan belas peratus untuk spending lebih dua puluh tujuh ribu lima ratus dua puluh tujuh ringgit sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 1:10 sampai 23:05, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul satu sepuluh sampai dua puluh tiga kosong lima, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 88.1% untuk spending lebih RM 97,148 sebulan."}, {"role": "assistant", "content": "Cash back lapan puluh lapan perpuluhan satu peratus untuk spending lebih sembilan puluh tujuh ribu seratus empat puluh lapan ringgit sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 93% untuk spending lebih RM2,442 sebulan."}, {"role": "assistant", "content": "Cash back sembilan puluh tiga peratus untuk spending lebih dua ribu empat ratus empat puluh dua ringgit sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 78% untuk spending lebih RM994.72 sebulan."}, {"role": "assistant", "content": "Cash back tujuh puluh lapan peratus untuk spending lebih sembilan ratus sembilan puluh empat ringgit tujuh puluh dua sen sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 2% untuk spending lebih RM485 sebulan."}, {"role": "assistant", "content": "Cash back dua peratus untuk spending lebih empat ratus lapan puluh lima ringgit sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cash back 99% untuk spending lebih RM 302 sebulan."}, {"role": "assistant", "content": "Cash back sembilan puluh sembilan peratus untuk spending lebih tiga ratus dua ringgit sebulan."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Malay with English mixed in. Malaysian Malay-English code-switching (Bahasa rojak). RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Malay part is spoken in Malay, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kami buka dari pukul 6:45 pagi sampai 5 pagi, close on public holiday."}, {"role": "assistant", "content": "Kami buka dari pukul enam empat puluh lima pagi sampai lima pagi, close on public holiday."}], "lang": "ms-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight MH370 departs at 1.20a.m., please check in awal."}, {"role": "assistant", "content": "Your flight MH three seven zero departs at one twenty a m, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight ABC1234 departs at 2.30p.m., please check in awal."}, {"role": "assistant", "content": "Your flight ABC one two three four departs at two thirty p m, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight TK-88213 departs at 12.20PM, please check in awal."}, {"role": "assistant", "content": "Your flight TK eight eight two one three departs at twelve twenty P M, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight QZ8501 departs at 2pm, please check in awal."}, {"role": "assistant", "content": "Your flight QZ eight five zero one departs at two p m, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight WXY 1234 departs at 17:40, please check in awal."}, {"role": "assistant", "content": "Your flight WXY one two three four departs at seventeen forty, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight REF-2024-0113 departs at 2.10AM, please check in awal."}, {"role": "assistant", "content": "Your flight REF two zero two four zero one one three departs at two ten A M, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight PO 55219 departs at 6AM, please check in awal."}, {"role": "assistant", "content": "Your flight PO five five two one nine departs at six A M, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight TK-88213 departs at 10:05 AM, please check in awal."}, {"role": "assistant", "content": "Your flight TK eight eight two one three departs at ten zero five A M, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight INV0472 departs at 2.00p.m., please check in awal."}, {"role": "assistant", "content": "Your flight INV zero four seven two departs at two p m, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight A12 departs at 1:40 AM, please check in awal."}, {"role": "assistant", "content": "Your flight A twelve departs at one forty A M, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight B2-14 departs at 2AM, please check in awal."}, {"role": "assistant", "content": "Your flight B two fourteen departs at two A M, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight X-12340567 departs at 10p.m., please check in awal."}, {"role": "assistant", "content": "Your flight X one two three four zero five six seven departs at ten p m, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight REF-2024-0113 departs at 8:20, please check in awal."}, {"role": "assistant", "content": "Your flight REF two zero two four zero one one three departs at eight twenty, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight PO 55219 departs at 11:45 AM, please check in awal."}, {"role": "assistant", "content": "Your flight PO five five two one nine departs at eleven forty-five A M, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight PO 55219 departs at 9.10pm, please check in awal."}, {"role": "assistant", "content": "Your flight PO five five two one nine departs at nine ten p m, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight INV0472 departs at 11:30 pm, please check in awal."}, {"role": "assistant", "content": "Your flight INV zero four seven two departs at eleven thirty p m, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight MH370 departs at 1.20am, please check in awal."}, {"role": "assistant", "content": "Your flight MH three seven zero departs at one twenty a m, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight X-12340567 departs at 2:50 p.m., please check in awal."}, {"role": "assistant", "content": "Your flight X one two three four zero five six seven departs at two fifty p m, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight X-12340567 departs at 11:05 p.m., please check in awal."}, {"role": "assistant", "content": "Your flight X one two three four zero five six seven departs at eleven zero five p m, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight MH370 departs at 1.00am, please check in awal."}, {"role": "assistant", "content": "Your flight MH three seven zero departs at one a m, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight A12 departs at 3.10p.m., please check in awal."}, {"role": "assistant", "content": "Your flight A twelve departs at three ten p m, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight A12 departs at 1.40AM, please check in awal."}, {"role": "assistant", "content": "Your flight A twelve departs at one forty A M, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight WXY 1234 departs at 21:30, please check in awal."}, {"role": "assistant", "content": "Your flight WXY one two three four departs at twenty-one thirty, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight WXY 1234 departs at 5.15PM, please check in awal."}, {"role": "assistant", "content": "Your flight WXY one two three four departs at five fifteen P M, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight AK6120 departs at 12AM, please check in awal."}, {"role": "assistant", "content": "Your flight AK six one two zero departs at twelve A M, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight REF-2024-0113 departs at 9.20AM, please check in awal."}, {"role": "assistant", "content": "Your flight REF two zero two four zero one one three departs at nine twenty A M, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight INV0472 departs at 12:20 AM, please check in awal."}, {"role": "assistant", "content": "Your flight INV zero four seven two departs at twelve twenty A M, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight TK-88213 departs at 10.00a.m., please check in awal."}, {"role": "assistant", "content": "Your flight TK eight eight two one three departs at ten a m, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight AK6120 departs at 7a.m., please check in awal."}, {"role": "assistant", "content": "Your flight AK six one two zero departs at seven a m, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight AK6120 departs at 12:15, please check in awal."}, {"role": "assistant", "content": "Your flight AK six one two zero departs at twelve fifteen, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight A12 departs at 12:05 AM, please check in awal."}, {"role": "assistant", "content": "Your flight A twelve departs at twelve zero five A M, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight AK6120 departs at 6PM, please check in awal."}, {"role": "assistant", "content": "Your flight AK six one two zero departs at six P M, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight INV0472 departs at 8pm, please check in awal."}, {"role": "assistant", "content": "Your flight INV zero four seven two departs at eight p m, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight TK-88213 departs at 4:10 AM, please check in awal."}, {"role": "assistant", "content": "Your flight TK eight eight two one three departs at four ten A M, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight REF-2024-0113 departs at 14:50, please check in awal."}, {"role": "assistant", "content": "Your flight REF two zero two four zero one one three departs at fourteen fifty, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight B2-14 departs at 7AM, please check in awal."}, {"role": "assistant", "content": "Your flight B two fourteen departs at seven A M, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight INV0472 departs at 6:30 p.m., please check in awal."}, {"role": "assistant", "content": "Your flight INV zero four seven two departs at six thirty p m, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight MH370 departs at 12.40am, please check in awal."}, {"role": "assistant", "content": "Your flight MH three seven zero departs at twelve forty a m, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight AK6120 departs at 8.30am, please check in awal."}, {"role": "assistant", "content": "Your flight AK six one two zero departs at eight thirty a m, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight QZ8501 departs at 13:40, please check in awal."}, {"role": "assistant", "content": "Your flight QZ eight five zero one departs at thirteen forty, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight A12 departs at 7.50a.m., please check in awal."}, {"role": "assistant", "content": "Your flight A twelve departs at seven fifty a m, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight MH370 departs at 8:30, please check in awal."}, {"role": "assistant", "content": "Your flight MH three seven zero departs at eight thirty, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight ABC1234 departs at 2a.m., please check in awal."}, {"role": "assistant", "content": "Your flight ABC one two three four departs at two a m, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight AK6120 departs at 11:30 pm, please check in awal."}, {"role": "assistant", "content": "Your flight AK six one two zero departs at eleven thirty p m, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight ABC1234 departs at 9:40 PM, please check in awal."}, {"role": "assistant", "content": "Your flight ABC one two three four departs at nine forty P M, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight ABC1234 departs at 9pm, please check in awal."}, {"role": "assistant", "content": "Your flight ABC one two three four departs at nine p m, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight B2-14 departs at 5:20 am, please check in awal."}, {"role": "assistant", "content": "Your flight B two fourteen departs at five twenty a m, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight X-12340567 departs at 2:10, please check in awal."}, {"role": "assistant", "content": "Your flight X one two three four zero five six seven departs at two ten, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight QZ8501 departs at 12PM, please check in awal."}, {"role": "assistant", "content": "Your flight QZ eight five zero one departs at twelve P M, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight A12 departs at 12.00pm, please check in awal."}, {"role": "assistant", "content": "Your flight A twelve departs at twelve p m, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight PO 55219 departs at 6:10 am, please check in awal."}, {"role": "assistant", "content": "Your flight PO five five two one nine departs at six ten a m, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly English with Malay mixed in. Malaysian English with Malay mixed in. RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the English part is spoken in English, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your flight PO 55219 departs at 10:05 AM, please check in awal."}, {"role": "assistant", "content": "Your flight PO five five two one nine departs at ten zero five A M, please check in awal."}], "lang": "en-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM306就有16%的cash back。"}, {"role": "assistant", "content": "消费满三百零六令吉就有百分之十六的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM698.50就有64.8%的cash back。"}, {"role": "assistant", "content": "消费满六百九十八令吉五十仙就有百分之六十四点八的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM 89.94就有13%的cash back。"}, {"role": "assistant", "content": "消费满八十九令吉九十四仙就有百分之十三的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM 40.41就有65%的cash back。"}, {"role": "assistant", "content": "消费满四十令吉四十一仙就有百分之六十五的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM850就有56.7%的cash back。"}, {"role": "assistant", "content": "消费满八百五十令吉就有百分之五十六点七的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM 46.90就有5.8%的cash back。"}, {"role": "assistant", "content": "消费满四十六令吉九十仙就有百分之五点八的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM220.26就有71%的cash back。"}, {"role": "assistant", "content": "消费满二百二十令吉二十六仙就有百分之七十一的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM13,724就有22.7%的cash back。"}, {"role": "assistant", "content": "消费满一万三千七百二十四令吉就有百分之二十二点七的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM45,108就有46.4%的cash back。"}, {"role": "assistant", "content": "消费满四万五千一百零八令吉就有百分之四十六点四的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM51.57就有51%的cash back。"}, {"role": "assistant", "content": "消费满五十一令吉五十七仙就有百分之五十一的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM 40.90就有44.6%的cash back。"}, {"role": "assistant", "content": "消费满四十令吉九十仙就有百分之四十四点六的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM 889.90就有87.7%的cash back。"}, {"role": "assistant", "content": "消费满八百八十九令吉九十仙就有百分之八十七点七的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM 787.50就有87%的cash back。"}, {"role": "assistant", "content": "消费满七百八十七令吉五十仙就有百分之八十七的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM21.50就有1%的cash back。"}, {"role": "assistant", "content": "消费满二十一令吉五十仙就有百分之一的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM 284就有51.9%的cash back。"}, {"role": "assistant", "content": "消费满二百八十四令吉就有百分之五十一点九的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM27就有24%的cash back。"}, {"role": "assistant", "content": "消费满二十七令吉就有百分之二十四的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM 57.86就有33.7%的cash back。"}, {"role": "assistant", "content": "消费满五十七令吉八十六仙就有百分之三十三点七的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM24.50就有80%的cash back。"}, {"role": "assistant", "content": "消费满二十四令吉五十仙就有百分之八十的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM249.89就有2%的cash back。"}, {"role": "assistant", "content": "消费满二百四十九令吉八十九仙就有百分之二的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM 97.90就有45%的cash back。"}, {"role": "assistant", "content": "消费满九十七令吉九十仙就有百分之四十五的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM77,001.90就有8.5%的cash back。"}, {"role": "assistant", "content": "消费满七万七千零一令吉九十仙就有百分之八点五的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM 995就有69%的cash back。"}, {"role": "assistant", "content": "消费满九百九十五令吉就有百分之六十九的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM 636.74就有27.8%的cash back。"}, {"role": "assistant", "content": "消费满六百三十六令吉七十四仙就有百分之二十七点八的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM14,297就有25%的cash back。"}, {"role": "assistant", "content": "消费满一万四千二百九十七令吉就有百分之二十五的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM 69就有73%的cash back。"}, {"role": "assistant", "content": "消费满六十九令吉就有百分之七十三的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM38就有23%的cash back。"}, {"role": "assistant", "content": "消费满三十八令吉就有百分之二十三的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM769.50就有71.4%的cash back。"}, {"role": "assistant", "content": "消费满七百六十九令吉五十仙就有百分之七十一点四的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM 337.50就有12%的cash back。"}, {"role": "assistant", "content": "消费满三百三十七令吉五十仙就有百分之十二的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM994就有13.5%的cash back。"}, {"role": "assistant", "content": "消费满九百九十四令吉就有百分之十三点五的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM 548.90就有30%的cash back。"}, {"role": "assistant", "content": "消费满五百四十八令吉九十仙就有百分之三十的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM 15.07就有94%的cash back。"}, {"role": "assistant", "content": "消费满十五令吉七仙就有百分之九十四的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM76.79就有10%的cash back。"}, {"role": "assistant", "content": "消费满七十六令吉七十九仙就有百分之十的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM72,126.60就有67.9%的cash back。"}, {"role": "assistant", "content": "消费满七万二千一百二十六令吉六十仙就有百分之六十七点九的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM188就有66%的cash back。"}, {"role": "assistant", "content": "消费满一百八十八令吉就有百分之六十六的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM 35,071.50就有16.7%的cash back。"}, {"role": "assistant", "content": "消费满三万五千零七十一令吉五十仙就有百分之十六点七的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM 252.33就有52%的cash back。"}, {"role": "assistant", "content": "消费满二百五十二令吉三十三仙就有百分之五十二的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM 11,740就有77.6%的cash back。"}, {"role": "assistant", "content": "消费满一万一千七百四十令吉就有百分之七十七点六的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM3就有57%的cash back。"}, {"role": "assistant", "content": "消费满三令吉就有百分之五十七的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM29就有11%的cash back。"}, {"role": "assistant", "content": "消费满二十九令吉就有百分之十一的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with English mixed in. Malaysian Mandarin with English mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "消费满RM74,045.50就有54%的cash back。"}, {"role": "assistant", "content": "消费满七万四千零四十五令吉五十仙就有百分之五十四的cash back。"}], "lang": "zh-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor +60 14-610 1608，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 加六零一四六一零一六零八，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 03-2925 1583，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零三二九二五一五八三，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor +60 19-645 6448，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 加六零一九六四五六四四八，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 03-9459 5174，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零三九四五九五一七四，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 010-41109148，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零一零四一一零九一四八，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor +60 17-499 0288，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 加六零一七四九九零二八八，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor +60 11-757 3546，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 加六零一一七五七三五四六，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 014-42227907，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零一四四二二二七九零七，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 011-871 5430，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零一一八七一五四三零，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor +60 14-744 4683，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 加六零一四七四四四六八三，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 03-3256 1604，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零三三二五六一六零四，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 010-15316233，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零一零一五三一六二三三，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor +60 12-576 6603，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 加六零一二五七六六六零三，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 016-37350598，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零一六三七三五零五九八，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor +60 13-264 7736，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 加六零一三二六四七七三六，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 019-51870622，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零一九五一八七零六二二，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor +60 19-988 1653，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 加六零一九九八八一六五三，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 010-07185146，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零一零零七一八五一四六，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 018-049 0119，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零一八零四九零一一九，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 014-267 0358，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零一四二六七零三五八，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 03-9775 2531，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零三九七七五二五三一，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 019-846 5627，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零一九八四六五六二七，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 019-36024682，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零一九三六零二四六八二，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 017-19148981，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零一七一九一四八九八一，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 03-9724 9572，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零三九七二四九五七二，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 03-7071 9193，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零三七零七一九一九三，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 013-08897142，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零一三零八八九七一四二，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 03-1115 8967，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零三一一一五八九六七，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 03-0387 5110，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零三零三八七五一一零，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 010-37652764，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零一零三七六五二七六四，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor +60 16-298 5654，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 加六零一六二九八五六五四，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 014-014 9989，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零一四零一四九九八九，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 017-200 0166，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零一七二零零零一六六，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor +60 11-224 0808，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 加六零一一二二四零八零八，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 03-7931 2178，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零三七九三一二一七八，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 010-35693591，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零一零三五六九三五九一，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 012-18072341，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零一二一八零七二三四一，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor +60 19-144 9836，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 加六零一九一四四九八三六，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 03-4897 3338，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零三四八九七三三三八，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 03-9663 3469，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零三九六六三三四六九，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor +60 19-815 0647，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 加六零一九八一五零六四七，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor +60 17-901 9539，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 加六零一七九零一九五三九，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor +60 12-455 8151，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 加六零一二四五五八一五一，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 018-470 3605，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零一八四七零三六零五，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 03-6824 4315，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零三六八二四四三一五，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor +60 11-916 5057，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 加六零一一九一六五零五七，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 03-5370 0280，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零三五三七零零二八零，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 013-23077090，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零一三二三零七七零九零，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 018-40413492，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零一八四零四一三四九二，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor +60 18-273 0815，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 加六零一八二七三零八一五，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 016-25491241，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零一六二五四九一二四一，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 03-9451 3920，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零三九四五一三九二零，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 03-2405 1826，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零三二四零五一八二六，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 03-5362 1797，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零三五三六二一七九七，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 03-7765 0334，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零三七七六五零三三四，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor +60 13-355 3288，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 加六零一三三五五三二八八，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 03-4980 2249，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零三四九八零二二四九，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor +60 11-839 7882，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 加六零一一八三九七八八二，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 019-40478852，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零一九四零四七八八五二，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor +60 11-984 0248，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 加六零一一九八四零二四八，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 03-7738 3999，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零三七七三八三九九九，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 03-3714 0593，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零三三七一四零五九三，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 03-3759 0690，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零三三七五九零六九零，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 03-5883 5315，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零三五八八三五三一五，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor +60 18-872 7849，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 加六零一八八七二七八四九，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor +60 13-547 0482，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 加六零一三五四七零四八二，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 015-349 2511，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零一五三四九二五一一，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor +60 12-752 2478，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 加六零一二七五二二四七八，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 016-535 9222，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零一六五三五九二二二，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 03-2603 4855，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零三二六零三四八五五，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 018-28678951，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零一八二八六七八九五一，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 012-425 1436，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零一二四二五一四三六，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 03-6250 9299，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零三六二五零九二九九，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Mandarin Chinese with Malay mixed in. Malaysian Mandarin with Malay mixed in. RM is 令吉, cents 仙.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Mandarin Chinese part is spoken in Mandarin Chinese, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们会call您的nombor 010-523 1247，请保持畅通。"}, {"role": "assistant", "content": "我们会call您的nombor 零一零五二三一二四七，请保持畅通。"}], "lang": "zh-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM59 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் ஐம்பத்தொன்பது ரிங்கிட் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 34,461.50 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் முப்பத்து நான்காயிரத்து நானூற்று அறுபத்தொன்று ரிங்கிட் ஐம்பது சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 95,562.42 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் தொண்ணூற்று ஐந்தாயிரத்து ஐந்நூற்று அறுபத்திரண்டு ரிங்கிட் நாற்பத்திரண்டு சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM889.50 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் எண்ணூற்று எண்பத்தொன்பது ரிங்கிட் ஐம்பது சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM75,543 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் எழுபத்தைந்தாயிரத்து ஐந்நூற்று நாற்பத்து மூன்று ரிங்கிட் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM33.35 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் முப்பத்து மூன்று ரிங்கிட் முப்பத்தைந்து சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM30,702.90 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் முப்பதாயிரத்து எழுநூற்று இரண்டு ரிங்கிட் தொண்ணூறு சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM275 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் இருநூற்று எழுபத்தைந்து ரிங்கிட் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 730 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் எழுநூற்று முப்பது ரிங்கிட் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM79,379.96 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் எழுபத்தொன்பதாயிரத்து முந்நூற்று எழுபத்தொன்பது ரிங்கிட் தொண்ணூற்று ஆறு சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 996 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் தொள்ளாயிரத்து தொண்ணூற்று ஆறு ரிங்கிட் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM66,302.90 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் அறுபத்தாறாயிரத்து முந்நூற்று இரண்டு ரிங்கிட் தொண்ணூறு சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM784.03 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் எழுநூற்று எண்பத்து நான்கு ரிங்கிட் மூன்று சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM95 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் தொண்ணூற்று ஐந்து ரிங்கிட் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 84.33 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் எண்பத்து நான்கு ரிங்கிட் முப்பத்து மூன்று சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 85.90 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் எண்பத்தைந்து ரிங்கிட் தொண்ணூறு சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM3,922.90 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் மூவாயிரத்து தொள்ளாயிரத்து இருபத்திரண்டு ரிங்கிட் தொண்ணூறு சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM120.50 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் நூற்று இருபது ரிங்கிட் ஐம்பது சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 841.50 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் எண்ணூற்று நாற்பத்தொன்று ரிங்கிட் ஐம்பது சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 71,825 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் எழுபத்தொன்றாயிரத்து எண்ணூற்று இருபத்தைந்து ரிங்கிட் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM39.05 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் முப்பத்தொன்பது ரிங்கிட் ஐந்து சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 474 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் நானூற்று எழுபத்து நான்கு ரிங்கிட் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM28,660.90 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் இருபத்தெட்டாயிரத்து அறுநூற்று அறுபது ரிங்கிட் தொண்ணூறு சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 81,854 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் எண்பத்தொன்றாயிரத்து எண்ணூற்று ஐம்பத்து நான்கு ரிங்கிட் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM37 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் முப்பத்தேழு ரிங்கிட் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM606.50 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் அறுநூற்று ஆறு ரிங்கிட் ஐம்பது சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 405.90 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் நானூற்று ஐந்து ரிங்கிட் தொண்ணூறு சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 95,678.50 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் தொண்ணூற்று ஐந்தாயிரத்து அறுநூற்று எழுபத்தெட்டு ரிங்கிட் ஐம்பது சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 25.90 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் இருபத்தைந்து ரிங்கிட் தொண்ணூறு சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 37,228.77 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் முப்பத்தேழாயிரத்து இருநூற்று இருபத்தெட்டு ரிங்கிட் எழுபத்தேழு சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 99 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் தொண்ணூற்று ஒன்பது ரிங்கிட் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM447.21 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் நானூற்று நாற்பத்தேழு ரிங்கிட் இருபத்தொன்று சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM40,482.90 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் நாற்பதாயிரத்து நானூற்று எண்பத்திரண்டு ரிங்கிட் தொண்ணூறு சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 66 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் அறுபத்தாறு ரிங்கிட் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM43,350 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் நாற்பத்து மூன்றாயிரத்து முந்நூற்று ஐம்பது ரிங்கிட் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 19,486 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் பத்தொன்பதாயிரத்து நானூற்று எண்பத்தாறு ரிங்கிட் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM42.50 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் நாற்பத்திரண்டு ரிங்கிட் ஐம்பது சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM71,987.08 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் எழுபத்தொன்றாயிரத்து தொள்ளாயிரத்து எண்பத்தேழு ரிங்கிட் எட்டு சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM884 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் எண்ணூற்று எண்பத்து நான்கு ரிங்கிட் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM81,942.90 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் எண்பத்தொன்றாயிரத்து தொள்ளாயிரத்து நாற்பத்திரண்டு ரிங்கிட் தொண்ணூறு சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 70 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் எழுபது ரிங்கிட் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 89.90 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் எண்பத்தொன்பது ரிங்கிட் தொண்ணூறு சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 86.50 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் எண்பத்தாறு ரிங்கிட் ஐம்பது சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 55 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் ஐம்பத்தைந்து ரிங்கிட் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 46 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் நாற்பத்தாறு ரிங்கிட் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 89,132.90 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் எண்பத்தொன்பதாயிரத்து நூற்று முப்பத்திரண்டு ரிங்கிட் தொண்ணூறு சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 34.26 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் முப்பத்து நான்கு ரிங்கிட் இருபத்தாறு சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 5.90 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் ஐந்து ரிங்கிட் தொண்ணூறு சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 92.47 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் தொண்ணூற்று இரண்டு ரிங்கிட் நாற்பத்தேழு சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 64,423 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் அறுபத்து நான்காயிரத்து நானூற்று இருபத்து மூன்று ரிங்கிட் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 97,892.93 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் தொண்ணூற்று ஏழாயிரத்து எண்ணூற்று தொண்ணூற்று இரண்டு ரிங்கிட் தொண்ணூற்று மூன்று சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 303 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் முந்நூற்று மூன்று ரிங்கிட் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 42.90 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் நாற்பத்திரண்டு ரிங்கிட் தொண்ணூறு சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM8,890 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் எண்ணாயிரத்து எண்ணூற்று தொண்ணூறு ரிங்கிட் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 594.50 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் ஐந்நூற்று தொண்ணூற்று நான்கு ரிங்கிட் ஐம்பது சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 37,172 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் முப்பத்தேழாயிரத்து நூற்று எழுபத்திரண்டு ரிங்கிட் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 98 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் தொண்ணூற்று எட்டு ரிங்கிட் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM83,899 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் எண்பத்து மூன்றாயிரத்து எண்ணூற்று தொண்ணூற்று ஒன்பது ரிங்கிட் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM42,123 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் நாற்பத்திரண்டாயிரத்து நூற்று இருபத்து மூன்று ரிங்கிட் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM10.75 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் பத்து ரிங்கிட் எழுபத்தைந்து சென் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with English mixed in. Malaysian Tamil with English mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the English part in English. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் account-ல் RM 901 மட்டுமே உள்ளது, top up செய்யுங்கள்."}, {"role": "assistant", "content": "உங்கள் account-ல் தொள்ளாயிரத்து ஒன்று ரிங்கிட் மட்டுமே உள்ளது, top up செய்யுங்கள்."}], "lang": "ta-en", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 339 GB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data முந்நூற்று முப்பத்தொன்பது ஜிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 2.4Mbps பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data இரண்டு புள்ளி நான்கு எம்பிபிஎஸ் பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 8.9 GB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data எட்டு புள்ளி ஒன்பது ஜிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 7.3 GB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data ஏழு புள்ளி மூன்று ஜிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 483 MB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data நானூற்று எண்பத்து மூன்று எம்பி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 5.2TB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data ஐந்து புள்ளி இரண்டு டிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 73TB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data எழுபத்து மூன்று டிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 4.9GB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data நான்கு புள்ளி ஒன்பது ஜிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 6.5 Mbps பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data ஆறு புள்ளி ஐந்து எம்பிபிஎஸ் பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 38 MB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data முப்பத்தெட்டு எம்பி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 288Mbps பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data இருநூற்று எண்பத்தெட்டு எம்பிபிஎஸ் பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 5.4 Mbps பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data ஐந்து புள்ளி நான்கு எம்பிபிஎஸ் பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 5.3 GB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data ஐந்து புள்ளி மூன்று ஜிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 20TB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data இருபது டிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 170GB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data நூற்று எழுபது ஜிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 207 GB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data இருநூற்று ஏழு ஜிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 230MB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data இருநூற்று முப்பது எம்பி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 442 GB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data நானூற்று நாற்பத்திரண்டு ஜிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 9.4 Mbps பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data ஒன்பது புள்ளி நான்கு எம்பிபிஎஸ் பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 155GB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data நூற்று ஐம்பத்தைந்து ஜிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 4.5 GB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data நான்கு புள்ளி ஐந்து ஜிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 6.3TB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data ஆறு புள்ளி மூன்று டிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 9.2GB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data ஒன்பது புள்ளி இரண்டு ஜிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 7.1GB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data ஏழு புள்ளி ஒன்று ஜிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 3.5GB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data மூன்று புள்ளி ஐந்து ஜிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 134Mbps பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data நூற்று முப்பத்து நான்கு எம்பிபிஎஸ் பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 5.5Mbps பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data ஐந்து புள்ளி ஐந்து எம்பிபிஎஸ் பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 208 TB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data இருநூற்று எட்டு டிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 345GB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data முந்நூற்று நாற்பத்தைந்து ஜிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 148GB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data நூற்று நாற்பத்தெட்டு ஜிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 90 Mbps பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data தொண்ணூறு எம்பிபிஎஸ் பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 438GB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data நானூற்று முப்பத்தெட்டு ஜிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 222 GB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data இருநூற்று இருபத்திரண்டு ஜிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 483GB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data நானூற்று எண்பத்து மூன்று ஜிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 2.7 TB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data இரண்டு புள்ளி ஏழு டிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 7.6MB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data ஏழு புள்ளி ஆறு எம்பி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 391 MB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data முந்நூற்று தொண்ணூற்று ஒன்று எம்பி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 3.7Mbps பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data மூன்று புள்ளி ஏழு எம்பிபிஎஸ் பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 346 TB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data முந்நூற்று நாற்பத்தாறு டிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 417MB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data நானூற்று பதினேழு எம்பி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 216GB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data இருநூற்று பதினாறு ஜிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 4.1TB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data நான்கு புள்ளி ஒன்று டிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 17Mbps பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data பதினேழு எம்பிபிஎஸ் பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 110 TB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data நூற்று பத்து டிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 372GB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data முந்நூற்று எழுபத்திரண்டு ஜிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 6.2 MB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data ஆறு புள்ளி இரண்டு எம்பி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 10 Mbps பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data பத்து எம்பிபிஎஸ் பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 257GB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data இருநூற்று ஐம்பத்தேழு ஜிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 99TB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data தொண்ணூற்று ஒன்பது டிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 327GB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data முந்நூற்று இருபத்தேழு ஜிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 2.4 MB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data இரண்டு புள்ளி நான்கு எம்பி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 79 GB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data எழுபத்தொன்பது ஜிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 9.4 GB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data ஒன்பது புள்ளி நான்கு ஜிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 131 MB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data நூற்று முப்பத்தொன்று எம்பி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 120MB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data நூற்று இருபது எம்பி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 353 GB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data முந்நூற்று ஐம்பத்து மூன்று ஜிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 124MB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data நூற்று இருபத்து நான்கு எம்பி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 20 Mbps பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data இருபது எம்பிபிஎஸ் பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 328 Mbps பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data முந்நூற்று இருபத்தெட்டு எம்பிபிஎஸ் பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 1.7TB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data ஒரு புள்ளி ஏழு டிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 1.2 GB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data ஒரு புள்ளி இரண்டு ஜிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 3.3TB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data மூன்று புள்ளி மூன்று டிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 113 GB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data நூற்று பதிமூன்று ஜிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 202GB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data இருநூற்று இரண்டு ஜிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. The text is code-switched: mostly Tamil with Malay mixed in. Malaysian Tamil with Malay mixed in. RM is ரிங்கிட், cents சென்.\nRules:\n- Expand everything that is not speakable as written: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP numbers digit by digit, dates, times, percentages, decimals, ordinals, units, email addresses and URLs.\n- Read each number in the language of the words AROUND it: a number inside the Tamil part is spoken in Tamil, a number inside the Malay part in Malay. Do not switch the sentence to one language.\n- Never answer questions, never add, drop, reorder or translate words. Every word that is already speakable stays exactly as written, in the language it was written in, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த மாதம் data 283 TB பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}, {"role": "assistant", "content": "இந்த மாதம் data இருநூற்று எண்பத்து மூன்று டிபி பயன்படுத்தி விட்டீர்கள், kuota முடிந்தது."}], "lang": "ta-ms", "source": "template"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Please settle your outstanding balance of RM1,200.40 by Friday to avoid service interruption."}, {"role": "assistant", "content": "Please settle your outstanding balance of one thousand two hundred ringgit forty sen by Friday to avoid service interruption."}], "lang": "en", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "The blood test results show a glucose level of 5.6 mmol/L."}, {"role": "assistant", "content": "The blood test results show a glucose level of five point six millimoles per litre."}], "lang": "en", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your credit card payment of RM150.20 is due on 15/10."}, {"role": "assistant", "content": "Your credit card payment of ringgit one hundred fifty point two zero is due on fifteen slash ten."}], "lang": "en", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Service interruption is expected from 0100 to 0500 hours for system maintenance."}, {"role": "assistant", "content": "Service interruption is expected from zero one zero zero to zero five zero zero hours for system maintenance."}], "lang": "en", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your request for a new SIM card will be processed by the 15. of next month."}, {"role": "assistant", "content": "Your request for a new SIM card will be processed by the fifteenth of next month."}], "lang": "en", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "I need to update my KYC info because my last update was back in 2017."}, {"role": "assistant", "content": "I need to update my KYC info because my last update was back in twenty seventeen."}], "lang": "en", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pls check your latest bank statement at https://onlinebanking.maybank.com.my to verify the RM1,200.50 transfer."}, {"role": "assistant", "content": "Pls check your latest bank statement at h t t p s colon slash slash onlinebanking dot maybank dot com dot m y to verify the ringgit one thousand two hundred point five zero transfer."}], "lang": "en", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "My phone line 012-3456789 is still lagging since yesterday."}, {"role": "assistant", "content": "My phone line zero one two three four five six seven eight nine is still lagging since yesterday."}], "lang": "en", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Please update the medical record for MRN 55432-X immediately."}, {"role": "assistant", "content": "Please update the medical record for MRN five five four three two dash X immediately."}], "lang": "en", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your loan application ref #AX772 will be processed by 15 Oct at 5pm."}, {"role": "assistant", "content": "Your loan application ref number A X seven seven two will be processed by fifteen October at five p m."}], "lang": "en", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your insurance coverage is 90% for admissions after 1 Jan 2024."}, {"role": "assistant", "content": "Your insurance coverage is ninety percent for admissions after first January twenty twenty-four."}], "lang": "en", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Your account balance is now RM-12.50 due to the monthly service charge."}, {"role": "assistant", "content": "Your account balance is now minus twelve ringgit fifty sen due to the monthly service charge."}], "lang": "en", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in English, ready to be synthesized. Malaysian English: RM is ringgit, sen for cents.\nRules:\n- Expand everything that is not speakable as written into English words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in English even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "My portfolio is down -2.4% this quarter because of the market crash."}, {"role": "assistant", "content": "My portfolio is down minus two point four percent this quarter because of the market crash."}], "lang": "en", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Sila ambil 3 biji tablet ini setiap 6 jam."}, {"role": "assistant", "content": "Sila ambil tiga biji tablet ini setiap enam jam."}], "lang": "ms", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Sila pastikan baki kredit anda mencukupi sekurang-kurangnya RM10.00 untuk melanggan pelan data 50,000MB ini."}, {"role": "assistant", "content": "Sila pastikan baki kredit anda mencukupi sekurang-kurangnya sepuluh ringgit untuk melanggan pelan data lima puluh ribu megabait ini."}], "lang": "ms", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Baki akaun simpanan anda kini tinggal RM12.50 sahaja."}, {"role": "assistant", "content": "Baki akaun simpanan anda kini tinggal dua belas ringgit lima puluh sen sahaja."}], "lang": "ms", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Tolak 25% daripada jumlah bil bulanan anda dengan kod promo TELCO25."}, {"role": "assistant", "content": "Tolak dua puluh lima peratus daripada jumlah bil bulanan anda dengan kod promo TELCO dua puluh lima."}], "lang": "ms", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Bayaran RM120.40 untuk insurans kesihatan telah berjaya diproses."}, {"role": "assistant", "content": "Bayaran satu ratus dua puluh ringgit empat puluh sen untuk insurans kesihatan telah berjaya diproses."}], "lang": "ms", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Topup RM10 anda akan aktif secara automatik tepat pukul 12.00am ini."}, {"role": "assistant", "content": "Topup sepuluh ringgit anda akan aktif secara automatik tepat pukul dua belas titik sifar sifar pagi ini."}], "lang": "ms", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Bayaran bulanan RM89.90 akan ditolak dari akaun anda pada jam 00:00."}, {"role": "assistant", "content": "Bayaran bulanan lapan puluh sembilan ringgit sembilan puluh sen akan ditolak dari akaun anda pada jam kosong kosong kosong kosong."}], "lang": "ms", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kenaikan caj servis akan bermula pada suku ke-4 tahun ini."}, {"role": "assistant", "content": "Kenaikan caj servis akan bermula pada suku keempat tahun ini."}], "lang": "ms", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Waktu melawat pesakit di wad adalah dari 4pm-6pm sahaja."}, {"role": "assistant", "content": "Waktu melawat pesakit di wad adalah dari pukul empat petang hingga pukul enam petang sahaja."}], "lang": "ms", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Sila kurangkan pengambilan gula dalam diet harian kepada bawah 25 g."}, {"role": "assistant", "content": "Sila kurangkan pengambilan gula dalam diet harian kepada bawah dua puluh lima gram."}], "lang": "ms", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kos rawatan pembedahan itu mencecah RM12,400.50, jumlah tertinggi sejak 2010."}, {"role": "assistant", "content": "Kos rawatan pembedahan itu mencecah dua belas ribu empat ratus ringgit lima puluh sen, jumlah tertinggi sejak dua ribu sepuluh."}], "lang": "ms", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Transaksi berjumlah RM1,200.50 telah dipindahkan ke akaun akaun 55667788."}, {"role": "assistant", "content": "Transaksi berjumlah satu ribu dua ratus ringgit lima puluh sen telah dipindahkan ke akaun akaun lima lima enam enam tujuh tujuh lapan lapan."}], "lang": "ms", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Doktor kata dos ubat ni kena ambil 1/2 tablet sahaja sebelum tidur."}, {"role": "assistant", "content": "Doktor kata dos ubat ni kena ambil satu per dua tablet sahaja sebelum tidur."}], "lang": "ms", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Sila hantar borang tuntutan insurans dalam masa 1/2 hari bekerja."}, {"role": "assistant", "content": "Sila hantar borang tuntutan insurans dalam masa satu per dua hari bekerja."}], "lang": "ms", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Malay, ready to be synthesized. Bahasa Melayu Malaysia (RM = ringgit, sen).\nRules:\n- Expand everything that is not speakable as written into Malay words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Malay even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Suhu freezer kat lab ni dah cecah -80°C, jadi sampel selamat."}, {"role": "assistant", "content": "Suhu freezer kat lab ni dah cecah negatif lapan puluh darjah Celsius, jadi sampel selamat."}], "lang": "ms", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Cicilan bulanan kartu kredit Anda sebesar Rp750.250,50."}, {"role": "assistant", "content": "Cicilan bulanan kartu kredit Anda sebesar tujuh ratus lima puluh ribu dua ratus lima puluh rupiah koma lima puluh."}], "lang": "id", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Berat badan bayi naik 0,75 kg dalam satu bulan terakhir."}, {"role": "assistant", "content": "Berat badan bayi naik nol koma tujuh lima kilogram dalam satu bulan terakhir."}], "lang": "id", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Suku bunga KPR naik jadi 8,5% mulai bulan depan."}, {"role": "assistant", "content": "Suku bunga KPR naik jadi delapan koma lima persen mulai bulan depan."}], "lang": "id", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pembayaran iuran BPJS paling lambat tgl 10 setiap bulannya."}, {"role": "assistant", "content": "Pembayaran iuran BPJS paling lambat tanggal sepuluh setiap bulannya."}], "lang": "id", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Sinyal di area saya sering hilang kalau sudah lewat pukul 23:00."}, {"role": "assistant", "content": "Sinyal di area saya sering hilang kalau sudah lewat pukul dua puluh tiga nol nol."}], "lang": "id", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Antrean obat di apotek biasanya ramai sekali sekitar jam 1 siang."}, {"role": "assistant", "content": "Antrean obat di apotek biasanya ramai sekali sekitar jam satu siang."}], "lang": "id", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pasien diminta datang 15 menit sebelum jadwal operasi jam 10:00 pagi."}, {"role": "assistant", "content": "Pasien diminta datang lima belas menit sebelum jadwal operasi jam sepuluh nol nol pagi."}], "lang": "id", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Silakan hubungi call center di jam 09.00-21.00 kalau ada kendala sinyal."}, {"role": "assistant", "content": "Silakan hubungi call center di jam sembilan nol nol sampai dua puluh satu nol nol kalau ada kendala sinyal."}], "lang": "id", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paket data internet Anda akan berakhir, silakan isi ulang via aplikasi dengan kode promo NET2024."}, {"role": "assistant", "content": "Paket data internet Anda akan berakhir, silakan isi ulang via aplikasi dengan kode promo N E T dua ribu dua puluh empat."}], "lang": "id", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Indonesian, ready to be synthesized. Bahasa Indonesia (Rp = rupiah).\nRules:\n- Expand everything that is not speakable as written into Indonesian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Indonesian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Jadwal kontrol dokter spesialis jam 10.30 pagi, kode bookingnya HC-99."}, {"role": "assistant", "content": "Jadwal kontrol dokter spesialis jam sepuluh tiga puluh pagi, kode bookingnya H C sembilan sembilan."}], "lang": "id", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我的手机信号很差，已经断线 3 次了。"}, {"role": "assistant", "content": "我的手机信号很差，已经断线 三 次了。"}], "lang": "zh", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "由于网络故障，目前约 20% 的用户无法上网。"}, {"role": "assistant", "content": "由于网络故障，目前约 二十百分比 的用户无法上网。"}], "lang": "zh", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请在三个工作日内将 RM1,250.50 转账至指定的银行账户。"}, {"role": "assistant", "content": "请在三个工作日内将 一千二百五十点五零令吉 转账至指定的银行账户。"}], "lang": "zh", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "目前定期存款的年利率是 3.8%，最低存款额为 RM5,000。"}, {"role": "assistant", "content": "目前定期存款的年利率是 百分之三点八，最低存款额为 五千令吉。"}], "lang": "zh", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "如果您在网上办理开户，首年年费可享 100% 折扣。"}, {"role": "assistant", "content": "如果您在网上办理开户，首年年费可享 一百分之百 折扣。"}], "lang": "zh", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "目前的余额不足以支付令吉 50 的月费，请尽快充值。"}, {"role": "assistant", "content": "目前的余额不足以支付令吉 五十 的月费，请尽快充值。"}], "lang": "zh", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您的配套将在 24 小时内激活，如有疑问请拨打 100 或 011-12345678。"}, {"role": "assistant", "content": "您的配套将在 二十四 小时内激活，如有疑问请拨打 一零零 或 零一一-一二三四五六七八。"}], "lang": "zh", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您的OTP验证码是 8 2 0 4 1 9，请在5分钟内输入。"}, {"role": "assistant", "content": "您的OTP验证码是 八 二 零 四 一 九，请在五分钟内输入。"}], "lang": "zh", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "这个配套的第3个条款写得不清楚，请解释一下。"}, {"role": "assistant", "content": "这个配套的第三个条款写得不清楚，请解释一下。"}], "lang": "zh", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "请问 3rd floor 有没有 X-ray 部门？"}, {"role": "assistant", "content": "请问 第三层 有没有 X-ray 部门？"}], "lang": "zh", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "办理宽限期申请后，还款期限可延长1-3个月。"}, {"role": "assistant", "content": "办理宽限期申请后，还款期限可延长一到三个月。"}], "lang": "zh", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我的手机单欠款是从2023年起累积的，现在想分期付款。"}, {"role": "assistant", "content": "我的手机单欠款是从二零二三年起累积的，现在想分期付款。"}], "lang": "zh", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "医生说我这种病在1970s比较常见，现在已经很少见了。"}, {"role": "assistant", "content": "医生说我这种病在一九七零年代比较常见，现在已经很少见了。"}], "lang": "zh", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "预约医生请点击 https://www.kpjhealth.com.my 选择合适的时段。"}, {"role": "assistant", "content": "预约医生请点击 ｈｔｔｐｓ 冒号 斜杠 斜杠 ｗｗｗ 点 ｋｐｊｈｅａｌｔｈ 点 ｃｏｍ 点 ｍｙ 选择合适的时段。"}], "lang": "zh", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "您的手机号码 012-3456789 已成功激活 5G 网络服务。"}, {"role": "assistant", "content": "您的手机号码 零一二-三四五六七八九 已成功激活 五G 网络服务。"}], "lang": "zh", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "由于系统升级，请在 15/10/2023 之前通过 App 确认您的账户号 1029384756。"}, {"role": "assistant", "content": "由于系统升级，请在 二零二三年十月十五日 之前通过 App 确认您的账户号 一零二九三八四七五六。"}], "lang": "zh", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "我们的客户服务 24/7 全天候为您提供支持。"}, {"role": "assistant", "content": "我们的客户服务 二十四小时七天 全天候为您提供支持。"}], "lang": "zh", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "冷藏室的温度现在是-4.5度，请检查设备是否正常。"}, {"role": "assistant", "content": "冷藏室的温度现在是负四点五度，请检查设备是否正常。"}], "lang": "zh", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Mandarin Chinese, ready to be synthesized. Simplified Chinese as used in Malaysia (RM = 令吉).\nRules:\n- Expand everything that is not speakable as written into Mandarin Chinese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Mandarin Chinese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "由于系统错误，您的电子钱包被扣了-RM10.00，我们会尽快退款。"}, {"role": "assistant", "content": "由于系统错误，您的电子钱包被扣了负十令吉，我们会尽快退款。"}], "lang": "zh", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "என் அக்கவுண்டில் இருந்து RM 25.50 கட் ஆகிவிட்டது."}, {"role": "assistant", "content": "என் அக்கவுண்டில் இருந்து இருபத்தி ஐந்து ரிங்கிட் ஐம்பது காசுகள் கட் ஆகிவிட்டது."}], "lang": "ta", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்கள் கணக்கில் RM 15,000.00 மீதமுள்ளது."}, {"role": "assistant", "content": "உங்கள் கணக்கில் பதினைந்தாயிரம் ரிங்கிட் மீதமுள்ளது."}], "lang": "ta", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "டாக்டர் 5.5mg மாத்திரை சாப்பிட சொன்னாங்க."}, {"role": "assistant", "content": "டாக்டர் ஐந்து புள்ளி ஐந்து மில்லிகிராம் மாத்திரை சாப்பிட சொன்னாங்க."}], "lang": "ta", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "முதலீட்டின் மீது 6.25% லாபம் கிடைக்கும் என்று வங்கி உறுதி அளித்துள்ளது."}, {"role": "assistant", "content": "முதலீட்டின் மீது ஆறு புள்ளி இரண்டு ஐந்து சதவீதம் லாபம் கிடைக்கும் என்று வங்கி உறுதி அளித்துள்ளது."}], "lang": "ta", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "RM50 ரீலோட் 02/02/24 அன்று உங்கள் கணக்கில் சேர்க்கப்பட்டது."}, {"role": "assistant", "content": "ஐம்பது ரிங்கிட் ரீலோட் இரண்டு ஸ்லாஷ் இரண்டு ஸ்லாஷ் இருபத்தி நான்கு அன்று உங்கள் கணக்கில் சேர்க்கப்பட்டது."}], "lang": "ta", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிதாக இணைந்த வாடிக்கையாளர்களுக்கு முதல் 3-6 மாதங்கள் டேட்டா இலவசம்."}, {"role": "assistant", "content": "புதிதாக இணைந்த வாடிக்கையாளர்களுக்கு முதல் மூன்று முதல் ஆறு மாதங்கள் டேட்டா இலவசம்."}], "lang": "ta", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "புதிய சிம் கார்டு வாங்க RM 10 கட்டணம் செலுத்த வேண்டும்."}, {"role": "assistant", "content": "புதிய சிம் கார்டு வாங்க பத்து ரிங்கிட் கட்டணம் செலுத்த வேண்டும்."}], "lang": "ta", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "RM 2,500 இன்சூரன்ஸ் க்ளைம் செய்ய claim_dept@aia.com.my-க்கு டாக்குமெண்ட்ஸ் அனுப்பவும்."}, {"role": "assistant", "content": "இரண்டாயிரத்து ஐந்நூறு ரிங்கிட் இன்சூரன்ஸ் க்ளைம் செய்ய claim underscore dept at aia dot com dot my-க்கு டாக்குமெண்ட்ஸ் அனுப்பவும்."}], "lang": "ta", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil, ready to be synthesized. Tamil as used in Malaysia (RM = ரிங்கிட்).\nRules:\n- Expand everything that is not speakable as written into Tamil words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுடைய appointment slip-ல் உள்ள ref: APP-10293-B-ஐ வைத்து உறுதி செய்யவும்."}, {"role": "assistant", "content": "உங்களுடைய appointment slip-ல் உள்ள ref: ஏ பி பி ஒன்று பூஜ்ஜியம் இரண்டு ஒன்பது மூன்று பி-ஐ வைத்து உறுதி செய்யவும்."}], "lang": "ta", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுடைய அக்ரவுண்ட் நம்பர் 1-0-0-2-3-4-5-6-7-8-9-0-1-2-3-4-5-6-7-8-9."}, {"role": "assistant", "content": "உங்களுடைய அக்ரவுண்ட் நம்பர் ஒன்று பூஜ்ஜியம் பூஜ்ஜியம் இரண்டு மூன்று நான்கு ஐந்து ஆறு ஏழு எட்டு ஒன்பது பூஜ்ஜியம் ஒன்று இரண்டு மூன்று நான்கு ஐந்து ஆறு ஏழு எட்டு ஒன்பது."}], "lang": "ta-LK", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "கடன் விண்ணப்பத்தை சமர்ப்பிக்க இன்று மாலை 4:00 மணிக்கெல்லாம் வரவும்."}, {"role": "assistant", "content": "கடன் விண்ணப்பத்தை சமர்ப்பிக்க இன்று மாலை நான்கு மணி மணிக்கெல்லாம் வரவும்."}], "lang": "ta-LK", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இலங்கை தொலைத்தொடர்பு நிறுவனத்தின் ரீலோட் 50-100 ரூபாய் வரை கிடைக்கும்."}, {"role": "assistant", "content": "இலங்கை தொலைத்தொடர்பு நிறுவனத்தின் ரீலோட் ஐம்பது முதல் நூறு ரூபாய் வரை கிடைக்கும்."}], "lang": "ta-LK", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இரத்த அழுத்தத்தைக் குறைக்க உப்பை 5 g க்கும் குறைவாகப் பயன்படுத்துங்கள்."}, {"role": "assistant", "content": "இரத்த அழுத்தத்தைக் குறைக்க உப்பை ஐந்து கிராம் க்கும் குறைவாகப் பயன்படுத்துங்கள்."}], "lang": "ta-LK", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த இன்சூரன்ஸ் திட்டத்தின் கீழ் 5 kg தங்கம் வரை பாதுகாப்பாக வைப்பீடு செய்ய முடியும்."}, {"role": "assistant", "content": "இந்த இன்சூரன்ஸ் திட்டத்தின் கீழ் ஐந்து கிலோகிராம் தங்கம் வரை பாதுகாப்பாக வைப்பீடு செய்ய முடியும்."}], "lang": "ta-LK", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுடைய 100 GB டேட்டா பேக் இன்று நள்ளிரவுடன் காலாவதியாகிவிடும்."}, {"role": "assistant", "content": "உங்களுடைய நூறு ஜி பி டேட்டா பேக் இன்று நள்ளிரவுடன் காலாவதியாகிவிடும்."}], "lang": "ta-LK", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "இந்த டெல்கோ நிறுவனம் 2005-ல் ஆரம்பித்தது, இப்போது இலங்கையில் முன்னணி நிறுவனமாக உள்ளது."}, {"role": "assistant", "content": "இந்த டெல்கோ நிறுவனம் இரண்டாயிரத்து ஐந்து-ல் ஆரம்பித்தது, இப்போது இலங்கையில் முன்னணி நிறுவனமாக உள்ளது."}], "lang": "ta-LK", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுடைய டேட்டா பேக் முடிந்துவிட்டது, புதிய பேக் ஒன்றை www.dialog.lk இல் ஆக்டிவேட் செய்யவும்."}, {"role": "assistant", "content": "உங்களுடைய டேட்டா பேக் முடிந்துவிட்டது, புதிய பேக் ஒன்றை w w w dot dialog dot lk இல் ஆக்டிவேட் செய்யவும்."}], "lang": "ta-LK", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Tamil (Sri Lanka), ready to be synthesized. Tamil as used in Sri Lanka (Rs = ரூபாய், சதம்).\nRules:\n- Expand everything that is not speakable as written into Tamil (Sri Lanka) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Tamil (Sri Lanka) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "உங்களுடைய இரத்தப் பரிசோதனை அறிக்கை Lab-Ref: 77621-X இல் பதிவேற்றப்பட்டுள்ளது."}, {"role": "assistant", "content": "உங்களுடைய இரத்தப் பரிசோதனை அறிக்கை Lab-Ref: ஏழு ஏழு ஆறு இரண்டு ஒன்று hyphen X இல் பதிவேற்றப்பட்டுள்ளது."}], "lang": "ta-LK", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "මෙම වසරේ සෞඛ්‍ය අරමුදලට රු. 33,000,000 ක මුදලක් වෙන් කර ඇත."}, {"role": "assistant", "content": "මෙම වසරේ සෞඛ්‍ය අරමුදලට රුපියල් කෝටි තුනයියි ක මුදලක් වෙන් කර ඇත."}], "lang": "si", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "මෙම ඖෂධය දිනකට 2.5ml බැගින් ලබාගන්න."}, {"role": "assistant", "content": "මෙම ඖෂධය දිනකට දෙක දශම පහ මිලිලීටර් බැගින් ලබාගන්න."}], "lang": "si", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබේ ඉතිරි කිරීම් ගිණුමට දැන් 12.5% ක වාර්ෂික පොලියක් හිමිවේ."}, {"role": "assistant", "content": "ඔබේ ඉතිරි කිරීම් ගිණුමට දැන් සියයට දොළහයි දශම පහක වාර්ෂික පොලියක් හිමිවේ."}], "lang": "si", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "බෙහෙත් ඇණවුම අංක 902183 සඳහා ගෙවීම රු. 1,250 කි."}, {"role": "assistant", "content": "බෙහෙත් ඇණවුම අංක නවය බින්දුව දෙක එක අට තුන සඳහා ගෙවීම රුපියල් එකදහස් දෙසිය පනහයි කි."}], "lang": "si", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබේ ගිණුමෙන් රු. 1,200ක් අඩු වූවේ 2023-10-25 දින 16:20 ටයි."}, {"role": "assistant", "content": "ඔබේ ගිණුමෙන් රුපියල් එක්දහස් දෙකසියයක් අඩු වූවේ දෙදහස් විසිතුන ඔක්තෝබර් විසිපහ දින දහසයයි විස්සටයි."}], "lang": "si", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "රෝගියාගේ 3º report එක තවමත් ලැබී නැත."}, {"role": "assistant", "content": "රෝගියාගේ තුන්වන report එක තවමත් ලැබී නැත."}], "lang": "si", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "බැංකුවේ ATM යන්ත්‍රය සිට 1.2 km දුරින් ඇති වෙළඳසැලට යාමට මම තීරණය කළා."}, {"role": "assistant", "content": "බැංකුවේ ATM යන්ත්‍රය සිට එකයි දශම දෙකයි කිලෝමීටර් දුරින් ඇති වෙළඳසැලට යාමට මම තීරණය කළා."}], "lang": "si", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "අපගේ බැංකුව 1998 සිට පාරිභෝගිකයන්ට උපරිම සේවාවක් ලබා දෙයි."}, {"role": "assistant", "content": "අපගේ බැංකුව එක්දහස් නවසිය අනූඅට සිට පාරිභෝගිකයන්ට උපරිම සේවාවක් ලබා දෙයි."}], "lang": "si", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ඔබේ ණය ගිණුමේ account no: 100299384-01 වෙත මුදල් එවන්න."}, {"role": "assistant", "content": "ඔබේ ණය ගිණුමේ account no: එක බින්දුව බින්දුව දෙක නව නව තුන අට හතර හිනය බින්දුව එක වෙත මුදල් එවන්න."}], "lang": "si", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Sinhala, ready to be synthesized. Sinhala (රු. = රුපියල්).\nRules:\n- Expand everything that is not speakable as written into Sinhala words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Sinhala even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "හෙට දින අලුයම 8:00 සිට 11:00 දක්වා රෝහලේ සායනය විවෘතයි."}, {"role": "assistant", "content": "හෙට දින අලුයම අට සිට එකොළහ දක්වා රෝහලේ සායනය විවෘතයි."}], "lang": "si", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Masyadong mabagal ang internet connection namin ngayon."}, {"role": "assistant", "content": "Masyadong mabagal ang internet connection namin ngayon."}], "lang": "tl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kailangan ko nang mag-load ng ₱50.50 para hindi maputol ang data ko."}, {"role": "assistant", "content": "Kailangan ko nang mag-load ng limampung piso at limampung sentimo para hindi maputol ang data ko."}], "lang": "tl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ang downpayment ay 25% ng total price ng sasakyan."}, {"role": "assistant", "content": "Ang downpayment ay dalawampu't limang porsyento ng total price ng sasakyan."}], "lang": "tl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Tumawag po kayo sa 02-8111-2222 para ma-schedule ang inyong appointment sa cardiologist."}, {"role": "assistant", "content": "Tumawag po kayo sa zero dalawa guhit walo isa isa isa guhit dalawa dalawa dalawa dalawa para ma-schedule ang inyong appointment sa cardiologist."}], "lang": "tl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kailangan ko po yung 12-digit account number niyo, start sa 1-1-0-5-6-7-8-9-0-1-2."}, {"role": "assistant", "content": "Kailangan ko po yung labindalawang-digit account number niyo, start sa isa-isa-zero-lima-anim-pito-walo-siyam-zero-isa-dalawa."}], "lang": "tl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ma'am, paki-type po yung OTP na 7-0-4-5 sa app."}, {"role": "assistant", "content": "Ma'am, paki-type po yung OTP na pito-sero-apat-lima sa app."}], "lang": "tl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakisettle po ang balance niyo before 11/15 para iwas penalty."}, {"role": "assistant", "content": "Pakisettle po ang balance niyo before nobiyente labinlima para iwas penalty."}], "lang": "tl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "May appointment po kayo sa clinic bukas ng 10:30 am."}, {"role": "assistant", "content": "May appointment po kayo sa clinic bukas ng alas-diyes at kalahati ng umaga."}], "lang": "tl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Punta po kayo sa 2nd floor ng clinic para sa X-ray."}, {"role": "assistant", "content": "Punta po kayo sa ikalawang floor ng clinic para sa X-ray."}], "lang": "tl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ang promo na ito ay may unlimited calls at texts for 7-15 days depende sa load."}, {"role": "assistant", "content": "Ang promo na ito ay may unlimited calls at texts for pito hanggang labinlimang days depende sa load."}], "lang": "tl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakisend na lang po ng valid ID sa support@bankophp.com para ma-activate na yung account niyo."}, {"role": "assistant", "content": "Pakisend na lang po ng valid ID sa support at bankophp tuldok com para ma-activate na yung account niyo."}], "lang": "tl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Paki-update po ang inyong contact info sa account-mgmt@telcom.ph para iwas disconnection."}, {"role": "assistant", "content": "Paki-update po ang inyong contact info sa account-mgmt at telcom tuldok ph para iwas disconnection."}], "lang": "tl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Contact us at service@globetel.ph kung hindi pa rin gumagana ang internet niyo."}, {"role": "assistant", "content": "Contact us at service at globetel dot ph kung hindi pa rin gumagana ang internet niyo."}], "lang": "tl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "May promo kami na 20% discount sa monthly plan kung mag-upgrade ka before Oct 15."}, {"role": "assistant", "content": "May promo kami na dalawampung porsyentong discount sa monthly plan kung mag-upgrade ka before Oktubre labinlima."}], "lang": "tl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Filipino (Tagalog), ready to be synthesized. Filipino/Tagalog (₱ = piso); Taglish is acceptable.\nRules:\n- Expand everything that is not speakable as written into Filipino (Tagalog) words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Filipino (Tagalog) even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "May charge na ₱25 per transaction basta lagpas ₱10k ang amount."}, {"role": "assistant", "content": "May charge na dalawampu't limang piso per transaction basta lagpas sampung libong piso ang amount."}], "lang": "tl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "يرجى إعادة تشغيل الهاتف لتفعيل خدمات الجيل الخامس"}, {"role": "assistant", "content": "يرجى إعادة تشغيل الهاتف لتفعيل خدمات الجيل الخامس"}], "lang": "ar", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "عزيزي العميل، باقتك الحالية تمنحك 1000 دقيقة مكالمات محلية."}, {"role": "assistant", "content": "عزيزي العميل، باقتك الحالية تمنحك ألف دقيقة مكالمات محلية."}], "lang": "ar", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تكلفة تحديث البنية التحتية لشبكة الجيل الخامس بلغت 12,000,000 جنيه"}, {"role": "assistant", "content": "تكلفة تحديث البنية التحتية لشبكة الجيل الخامس بلغت اثنا عشر مليون جنيه"}], "lang": "ar", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تكلفة الفحص الشامل في المستشفى هي 1,200 ج.م."}, {"role": "assistant", "content": "تكلفة الفحص الشامل في المستشفى هي ألف ومائتان جنيه مصري."}], "lang": "ar", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "يرجى سداد الرصيد المستحق وقدره 150.75 AED لتجنب انقطاع الخدمة."}, {"role": "assistant", "content": "يرجى سداد الرصيد المستحق وقدره مائة وخمسون فاصلة خمسة وسبعون درهم إماراتي لتجنب انقطاع الخدمة."}], "lang": "ar", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تم استلام دفعة بقيمة 10,000 Dirhams بتاريخ 3 سبتمبر 2023."}, {"role": "assistant", "content": "تم استلام دفعة بقيمة عشرة آلاف درهم بتاريخ الثالث من سبتمبر ألفين وثلاثة وعشرين."}], "lang": "ar", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "يرجى العلم أن فرع البنك سيغلق اليوم في تمام الساعة 3:30 عصراً"}, {"role": "assistant", "content": "يرجى العلم أن فرع البنك سيغلق اليوم في تمام الساعة الثالثة والثلاثين دقيقة عصراً"}], "lang": "ar", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تأكد من تحديث تطبيق البنك للإصدار الـ12 لتجنب الأعطال"}, {"role": "assistant", "content": "تأكد من تحديث تطبيق البنك للإصدار الثاني عشر لتجنب الأعطال"}], "lang": "ar", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تتوفر عروض باقات الإنترنت الجديدة بخصومات تتراوح بين 10-20% للمشتركين الجدد."}, {"role": "assistant", "content": "تتوفر عروض باقات الإنترنت الجديدة بخصومات تتراوح بين عشرة إلى عشرين بالمئة للمشتركين الجدد."}], "lang": "ar", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "ستصلك رسالة تفعيل الشريحة في غضون 24-48 ساعة من تقديم الطلب."}, {"role": "assistant", "content": "ستصلك رسالة تفعيل الشريحة في غضون أربع وعشرين إلى ثمان وأربعين ساعة من تقديم الطلب."}], "lang": "ar", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "تستغرق عملية مراجعة الملف الطبي من 7-10 أيام عمل قبل تحديد موعد الجراحة."}, {"role": "assistant", "content": "تستغرق عملية مراجعة الملف الطبي من سبعة إلى عشرة أيام عمل قبل تحديد موعد الجراحة."}], "lang": "ar", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Arabic, ready to be synthesized. Modern Standard Arabic; amounts may be in riyal, dirham, pound or dollar.\nRules:\n- Expand everything that is not speakable as written into Arabic words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Arabic even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "باقة الإنترنت الخاصة بك انتهت، لتجديد الاشتراك بـ 100 جنيه يرجى مراسلتنا على activate@telco.com.eg"}, {"role": "assistant", "content": "باقة الإنترنت الخاصة بك انتهت، لتجديد الاشتراك بـ مائة جنيه يرجى مراسلتنا على activate أت telco نقطة com نقطة eg"}], "lang": "ar", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Le budget alloué à la recherche sur ce vaccin est de 45 000 000 €."}, {"role": "assistant", "content": "Le budget alloué à la recherche sur ce vaccin est de quarante-cinq millions d'euros."}], "lang": "fr", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "On m'a dit que 12.5% des patients ne tolèrent pas ce traitement spécifique."}, {"role": "assistant", "content": "On m'a dit que douze virgule cinq pour cent des patients ne tolèrent pas ce traitement spécifique."}], "lang": "fr", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "On est à 100% de capacité d'accueil aux urgences ce soir, c'est l'enfer."}, {"role": "assistant", "content": "On est à cent pour cent de capacité d'accueil aux urgences ce soir, c'est l'enfer."}], "lang": "fr", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "L'échéance de votre crédit immobilier est prévue pour le 31/01/25."}, {"role": "assistant", "content": "L'échéance de votre crédit immobilier est prévue pour le trente et un janvier deux mille vingt-cinq."}], "lang": "fr", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Rendez-vous pris pour le 14 oct. 2024 à 15h30 chez le cardiologue."}, {"role": "assistant", "content": "Rendez-vous pris pour le quatorze octobre deux mille vingt-quatre à quinze heures trente chez le cardiologue."}], "lang": "fr", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Contactez-moi vite sur p.lefevre@icloud.com pour le blocage de ma carte bleue."}, {"role": "assistant", "content": "Contactez-moi vite sur p point lefevre arobase icloud point com pour le blocage de ma carte bleue."}], "lang": "fr", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "L'accès à vos plafonds de carte est dispo sur https://www.societegenerale.fr/app"}, {"role": "assistant", "content": "L'accès à vos plafonds de carte est dispo sur hache t t p s deux points slash slash www point societe generale point fr slash app"}], "lang": "fr", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Le résultat de l'examen sanguin pour l'échantillon LAB-5521 est négatif."}, {"role": "assistant", "content": "Le résultat de l'examen sanguin pour l'échantillon L A B tiret cinq cinq deux un est négatif."}], "lang": "fr", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in French, ready to be synthesized. French (France).\nRules:\n- Expand everything that is not speakable as written into French words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in French even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il fait -7°C dehors, n'oubliez pas vos gants."}, {"role": "assistant", "content": "Il fait moins sept degrés Celsius dehors, n'oubliez pas vos gants."}], "lang": "fr", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para desbloquear la cuenta necesita transferir al menos 2.500€ a la cuenta puente."}, {"role": "assistant", "content": "Para desbloquear la cuenta necesita transferir al menos dos mil quinientos euros a la cuenta puente."}], "lang": "es", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Me han cobrado 12,40€ de exceso de datos en la última factura del móvil."}, {"role": "assistant", "content": "Me han cobrado doce euros con cuarenta céntimos de exceso de datos en la última factura del móvil."}], "lang": "es", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "El límite de transferencia diaria para esta cuenta es de 5.000€."}, {"role": "assistant", "content": "El límite de transferencia diaria para esta cuenta es de cinco mil euros."}], "lang": "es", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "El copago de la receta es de 4,12€."}, {"role": "assistant", "content": "El copago de la receta es de cuatro euros con doce céntimos."}], "lang": "es", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "El interés nominal de tu cuenta de ahorro es del 2,15% TAE."}, {"role": "assistant", "content": "El interés nominal de tu cuenta de ahorro es del dos coma quince por ciento TAE."}], "lang": "es", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Me han cobrado 12,50€ por la gestión de la transferencia internacional."}, {"role": "assistant", "content": "Me han cobrado doce euros con cincuenta céntimos por la gestión de la transferencia internacional."}], "lang": "es", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "El 20% de los usuarios reportan interferencias en la señal 5G."}, {"role": "assistant", "content": "El veinte por ciento de los usuarios reportan interferencias en la señal cinco ge."}], "lang": "es", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Si quiere cambiar su plan de datos, llame al 600 555 444."}, {"role": "assistant", "content": "Si quiere cambiar su plan de datos, llame al seis cero cero cinco cinco cinco cuatro cuatro cuatro."}], "lang": "es", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Recibirá el nuevo router en su domicilio el próximo martes, 17 de oct."}, {"role": "assistant", "content": "Recibirá el nuevo router en su domicilio el próximo martes, diecisiete de octubre."}], "lang": "es", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Su saldo actual a fecha de hoy, 10-04-2024, es de 1.250,50€."}, {"role": "assistant", "content": "Su saldo actual a fecha de hoy, diez de abril de dos mil veinticuatro, es de mil doscientos cincuenta euros con cincuenta céntimos."}], "lang": "es", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Su domicilio está ubicado en la 4ª planta, puerta B."}, {"role": "assistant", "content": "Su domicilio está ubicado en la cuarta planta, puerta B."}], "lang": "es", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "El plazo de reembolso es de 3-5 días hábiles."}, {"role": "assistant", "content": "El plazo de reembolso es de tres a cinco días hábiles."}], "lang": "es", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "El interés oscila entre el 2-4%."}, {"role": "assistant", "content": "El interés oscila entre el dos y cuatro por ciento."}], "lang": "es", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Si quiere cambiar su plan de datos, escriba a bajas_clientes@telco.es indicando su número de teléfono."}, {"role": "assistant", "content": "Si quiere cambiar su plan de datos, escriba a bajas guion bajo clientes arroba telco punto es indicando su número de teléfono."}], "lang": "es", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "El acceso al área de cliente está bloqueado, escribir a ayuda@fibra-rapida.es para resetear la clave."}, {"role": "assistant", "content": "El acceso al área de cliente está bloqueado, escribir a ayuda arroba fibra guion rapida punto es para resetear la clave."}], "lang": "es", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Para consultar el saldo de su cuenta, acceda a https://www.mibanco.es/privado"}, {"role": "assistant", "content": "Para consultar el saldo de su cuenta, acceda a h t t p s dos puntos barra barra w w w punto mibanco punto es barra privado"}], "lang": "es", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Su cita médica es el lunes 12 de mayo a las 17:15, por favor traiga su tarjeta sanitaria."}, {"role": "assistant", "content": "Su cita médica es el lunes doce de mayo a las diecisiete quince, por favor traiga su tarjeta sanitaria."}], "lang": "es", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "El rendimiento de los bonos ha sufrido una variación del -2,1% este trimestre."}, {"role": "assistant", "content": "El rendimiento de los bonos ha sufrido una variación del menos dos coma uno por ciento este trimestre."}], "lang": "es", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Spanish, ready to be synthesized. Spanish (Spain).\nRules:\n- Expand everything that is not speakable as written into Spanish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Spanish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Su plan de datos ha experimentado una degradación del -10.5% en la velocidad de subida."}, {"role": "assistant", "content": "Su plan de datos ha experimentado una degradación del menos diez punto cinco por ciento en la velocidad de subida."}], "lang": "es", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Für den Kredit benötigen wir 3 Gehaltsnachweise."}, {"role": "assistant", "content": "Für den Kredit benötigen wir drei Gehaltsnachweise."}], "lang": "de", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Wir haben gestern über 33.000.000 Euro in neue Infrastruktur investiert."}, {"role": "assistant", "content": "Wir haben gestern über dreiunddreißig Millionen Euro in neue Infrastruktur investiert."}], "lang": "de", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Das aktuelle Guthaben auf dem Firmenkonto beläuft sich auf 1.250.000,50 €, inklusive der Zinserträge."}, {"role": "assistant", "content": "Das aktuelle Guthaben auf dem Firmenkonto beläuft sich auf eine Million zweihundertfünfzigtausend Euro und fünfzig Cent, inklusive der Zinserträge."}], "lang": "de", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die neue 5G-Ausbaustrategie umfasst Investitionen von ca. 450.000.000 €."}, {"role": "assistant", "content": "Die neue fünf G-Ausbaustrategie umfasst Investitionen von circa vierhundertfünfzig Millionen Euro."}], "lang": "de", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Im letzten Jahr wurden landesweit 3.400.000 Impfdosen verabreicht."}, {"role": "assistant", "content": "Im letzten Jahr wurden landesweit drei Millionen vierhunderttausend Impfdosen verabreicht."}], "lang": "de", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Temperatur des Patienten ist auf 38,7 °C gestiegen."}, {"role": "assistant", "content": "Die Temperatur des Patienten ist auf achtunddreißig Komma sieben Grad Celsius gestiegen."}], "lang": "de", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Kontostand ist zu 100% gedeckt, sodass die Lastschrift ausgeführt werden kann."}, {"role": "assistant", "content": "Ihr Kontostand ist zu einhundert Prozent gedeckt, sodass die Lastschrift ausgeführt werden kann."}], "lang": "de", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ihr Dispokredit wird zum 15.03.2024 automatisch neu berechnet."}, {"role": "assistant", "content": "Ihr Dispokredit wird zum fünfzehnten März zweitausendvierundzwanzig automatisch neu berechnet."}], "lang": "de", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Überweisung wird voraussichtlich bis morgen, 11.05., 16:00 Uhr verbucht sein."}, {"role": "assistant", "content": "Die Überweisung wird voraussichtlich bis morgen, elfter Mai, sechzehn Uhr verbucht sein."}], "lang": "de", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Das Online-Banking ist heute Nacht von 02:00 bis 04:00 Uhr wegen Wartungsarbeiten nicht verfügbar."}, {"role": "assistant", "content": "Das Online-Banking ist heute Nacht von zwei Uhr bis vier Uhr wegen Wartungsarbeiten nicht verfügbar."}], "lang": "de", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ich rufe an, weil mein 3. Versuch, die SIM-Karte zu aktivieren, gescheitert ist."}, {"role": "assistant", "content": "Ich rufe an, weil mein dritter Versuch, die SIM-Karte zu aktivieren, gescheitert ist."}], "lang": "de", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die 1. Hilfe ist in diesem Raum direkt neben dem Aufzug."}, {"role": "assistant", "content": "Die erste Hilfe ist in diesem Raum direkt neben dem Aufzug."}], "lang": "de", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Bitte nehmen Sie 500ml Wasser zu sich, bevor Sie die Tablette einnehmen."}, {"role": "assistant", "content": "Bitte nehmen Sie fünfhundert Milliliter Wasser zu sich, bevor Sie die Tablette einnehmen."}], "lang": "de", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Für den Kreditantrag 2024 benötigen wir aktuelle Gehaltsnachweise der letzten 3 Monate."}, {"role": "assistant", "content": "Für den Kreditantrag zwei null zwei vier benötigen wir aktuelle Gehaltsnachweise der letzten drei Monate."}], "lang": "de", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Mein Vertrag läuft bis Ende 2024, kann ich ihn vorzeitig kündigen?"}, {"role": "assistant", "content": "Mein Vertrag läuft bis Ende zweitausendvierundzwanzig, kann ich ihn vorzeitig kündigen?"}], "lang": "de", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Einführung von SMS in den 1980er Jahren hat die Kommunikation revolutioniert."}, {"role": "assistant", "content": "Die Einführung von SMS in den neunzehnhundertachtziger Jahren hat die Kommunikation revolutioniert."}], "lang": "de", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Patientin leidet seit 2019 an chronischen Rückenschmerzen."}, {"role": "assistant", "content": "Die Patientin leidet seit zweitausendneunzehn an chronischen Rückenschmerzen."}], "lang": "de", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Können Sie mir die Bestätigung für die Überweisung an m.schmidt82@web.de schicken?"}, {"role": "assistant", "content": "Können Sie mir die Bestätigung für die Überweisung an m Punkt schmidt acht zwei At web Punkt de schicken?"}], "lang": "de", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Patientenakte unter der ID PAT-55432 ist bereits digital hinterlegt."}, {"role": "assistant", "content": "Die Patientenakte unter der ID P A T Bindestrich fünf fünf vier drei zwei ist bereits digital hinterlegt."}], "lang": "de", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in German, ready to be synthesized. German (Germany).\nRules:\n- Expand everything that is not speakable as written into German words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in German even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Die Kreditwürdigkeit wird auf einer Skala von 0/100 bewertet."}, {"role": "assistant", "content": "Die Kreditwürdigkeit wird auf einer Skala von null von einhundert bewertet."}], "lang": "de", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La tua SIM ha consumato 50.000 MB di traffico dati nell'ultimo mese."}, {"role": "assistant", "content": "La tua SIM ha consumato cinquantamila megabyte di traffico dati nell'ultimo mese."}], "lang": "it", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Ho bisogno del numero della carta, i primi sei cifre sono 5-4-1-2-8-8."}, {"role": "assistant", "content": "Ho bisogno del numero della carta, i primi sei cifre sono cinque quattro uno due otto otto."}], "lang": "it", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il vaccino richiamo va fatto entro il 22 nov."}, {"role": "assistant", "content": "Il vaccino richiamo va fatto entro il ventidue novembre."}], "lang": "it", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "La 2ª rata del finanziamento scadrà il prossimo mese."}, {"role": "assistant", "content": "La seconda rata del finanziamento scadrà il prossimo mese."}], "lang": "it", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Dagli anni '80 l'approccio alla gestione del rischio bancario è cambiato radicalmente."}, {"role": "assistant", "content": "Dagli anni ottanta l'approccio alla gestione del rischio bancario è cambiato radicalmente."}], "lang": "it", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Mi hanno addebitato 45€ di canone il 12 ott, potete verificare?"}, {"role": "assistant", "content": "Mi hanno addebitato quarantacinque euro di canone il dodici ottobre, potete verificare?"}], "lang": "it", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il punteggio di affidabilità creditizia è di 88/100."}, {"role": "assistant", "content": "Il punteggio di affidabilità creditizia è di ottantotto su cento."}], "lang": "it", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il saldo del conto è sceso a -150,50€ a causa delle commissioni."}, {"role": "assistant", "content": "Il saldo del conto è sceso a meno centocinquanta virgola cinquanta euro a causa delle commissioni."}], "lang": "it", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Italian, ready to be synthesized. Italian.\nRules:\n- Expand everything that is not speakable as written into Italian words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Italian even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Il tasso di interesse per questo prodotto è sceso a -0,25%."}, {"role": "assistant", "content": "Il tasso di interesse per questo prodotto è sceso a meno zero virgola venticinque per cento."}], "lang": "it", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O saldo da conta corrente é de 1.250.000€."}, {"role": "assistant", "content": "O saldo da conta corrente é de um milhão duzentos e cinquenta mil euros."}], "lang": "pt", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O volume de dados consumido foi de 2.000.000 MB."}, {"role": "assistant", "content": "O volume de dados consumido foi de dois milhões de megabytes."}], "lang": "pt", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Tenho um débito direto de 45€ para a operadora."}, {"role": "assistant", "content": "Tenho um débito direto de quarenta e cinco euros para a operadora."}], "lang": "pt", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "O seguro de saúde cobre 80% das despesas de internamento."}, {"role": "assistant", "content": "O seguro de saúde cobre oitenta por cento das despesas de internamento."}], "lang": "pt", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pode confirmar o seu número de telemóvel para enviarmos o código de ativação? O meu é 912345678."}, {"role": "assistant", "content": "Pode confirmar o seu número de telemóvel para enviarmos o código de ativação? O meu é nove um dois três quatro cinco seis sete oito."}], "lang": "pt", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Estou na 1ª fila para a triagem, mas isto não anda."}, {"role": "assistant", "content": "Estou na primeira fila para a triagem, mas isto não anda."}], "lang": "pt", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "A 4ª chamada para o apoio técnico da NOS caiu sozinha."}, {"role": "assistant", "content": "A quarta chamada para o apoio técnico da NOS caiu sozinha."}], "lang": "pt", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Introduza o código CVV e o valor de 19,99€ para concluir a compra hoje."}, {"role": "assistant", "content": "Introduza o código CVV e o valor de dezanove vírgula noventa e nove euros para concluir a compra hoje."}], "lang": "pt", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Portuguese, ready to be synthesized. European Portuguese (also R$ real).\nRules:\n- Expand everything that is not speakable as written into Portuguese words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Portuguese even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "A transferência de 500€ foi efetuada em 12/10/2023 às 11h15 via MB Way."}, {"role": "assistant", "content": "A transferência de quinhentos euros foi efetuada em doze de outubro de dois mil e vinte e três às onze horas e quinze via MB Way."}], "lang": "pt", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Kunt u mij helpen bij het instellen van mijn nieuwe mobiele telefoon?"}, {"role": "assistant", "content": "Kunt u mij helpen bij het instellen van mijn nieuwe mobiele telefoon?"}], "lang": "nl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "U moet uw identiteitsbewijs meenemen naar de apotheek."}, {"role": "assistant", "content": "U moet uw identiteitsbewijs meenemen naar de apotheek."}], "lang": "nl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Er is een storing in het netwerk in uw regio."}, {"role": "assistant", "content": "Er is een storing in het netwerk in uw regio."}], "lang": "nl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "De zorgverzekeraar heeft vorig jaar ruim 12.000.000 claims verwerkt."}, {"role": "assistant", "content": "De zorgverzekeraar heeft vorig jaar ruim twaalf miljoen claims verwerkt."}], "lang": "nl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uw mobiele abonnement is verlengd, de maandelijkse kosten bedragen nu € 22,99 incl. btw."}, {"role": "assistant", "content": "Uw mobiele abonnement is verlengd, de maandelijkse kosten bedragen nu tweeëntwintig euro negenennegentig cent incl. btw."}], "lang": "nl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Alle kosten worden voor 100% vergoed door de aanvullende verzekering."}, {"role": "assistant", "content": "Alle kosten worden voor honderd procent vergoed door de aanvullende verzekering."}], "lang": "nl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Zou u de postcode 5611 GK kunnen controleren in het systeem?"}, {"role": "assistant", "content": "Zou u de postcode vijf zes één één G K kunnen controleren in het systeem?"}], "lang": "nl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "De apotheek is vandaag geopend tot 18.00 uur."}, {"role": "assistant", "content": "De apotheek is vandaag geopend tot achttien uur."}], "lang": "nl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "De patiënt moet voor de 3e keer een bloedtest laten doen."}, {"role": "assistant", "content": "De patiënt moet voor de derde keer een bloedtest laten doen."}], "lang": "nl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Neem drie keer per dag een dosis van 5 ml van de siroop."}, {"role": "assistant", "content": "Neem drie keer per dag een dosis van vijf milliliter van de siroop."}], "lang": "nl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Log in via https://www. DigiD.nl om je afspraken bij het ziekenhuis te beheren."}, {"role": "assistant", "content": "Log in via hteteps twee punt nul punt www punt DigiD punt nl om je afspraken bij het ziekenhuis te beheren."}], "lang": "nl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "De koorts is gezakt, maar de kamer is nog steeds te koud met -2°C buiten."}, {"role": "assistant", "content": "De koorts is gezakt, maar de kamer is nog steeds te koud met min twee graden Celsius buiten."}], "lang": "nl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Dutch, ready to be synthesized. Dutch (Netherlands).\nRules:\n- Expand everything that is not speakable as written into Dutch words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Dutch even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Het is echt ijskoud vandaag, de thermometer geeft -4,5°C aan."}, {"role": "assistant", "content": "Het is echt ijskoud vandaag, de thermometer geeft min vier komma vijf graden Celsius aan."}], "lang": "nl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Proszę o otwarcie nowego rachunku oszczędnościowego."}, {"role": "assistant", "content": "Proszę o otwarcie nowego rachunku oszczędnościowego."}], "lang": "pl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Moja faktura za internet jest znacznie wyższa niż zwykle."}, {"role": "assistant", "content": "Moja faktura za internet jest znacznie wyższa niż zwykle."}], "lang": "pl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Rząd przeznaczył 45.000.000 zł na doposażenie szpitali powiatowych w nowoczesny sprzęt."}, {"role": "assistant", "content": "Rząd przeznaczył czterdzieści pięć milionów złotych na doposażenie szpitali powiatowych w nowoczesny sprzęt."}], "lang": "pl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Koszt dodatkowej karty SIM to tylko 4,99 zł."}, {"role": "assistant", "content": "Koszt dodatkowej karty SIM to tylko cztery złote dziewięćdziesiąt dziewięć groszy."}], "lang": "pl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Czy może Pan podyktować numer zamówienia 55921-X?"}, {"role": "assistant", "content": "Czy może Pan podyktować numer zamówienia pięć pięć dziewięć dwa jeden X?"}], "lang": "pl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Uprzejmie informuję, że 15. dzień miesiąca to ostateczny termin opłaty abonamentu."}, {"role": "assistant", "content": "Uprzejmie informuję, że piętnasty dzień miesiąca to ostateczny termin opłaty abonamentu."}], "lang": "pl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pacjent został przyjęty na 1. oddział kardiologii wczoraj wieczorem."}, {"role": "assistant", "content": "Pacjent został przyjęty na pierwszy oddział kardiologii wczoraj wieczorem."}], "lang": "pl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Plan taryfowy obejmuje nielimitowane rozmowy w godzinach 22-6 rano."}, {"role": "assistant", "content": "Plan taryfowy obejmuje nielimitowane rozmowy w godzinach dwudziestej drugiej do szóstej rano."}], "lang": "pl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja faktura za internet została wysłana na maila: jan.kowalski123@gmail.com."}, {"role": "assistant", "content": "Twoja faktura za internet została wysłana na maila: jan kropka kowalski jeden dwa trzy małpa gmail kropka com."}], "lang": "pl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Potwierdzenie zmiany limitów karty wysłaliśmy na adres klienta: m.nowak@wp.pl."}, {"role": "assistant", "content": "Potwierdzenie zmiany limitów karty wysłaliśmy na adres klienta: m kropka nowak małpa wp kropka pl."}], "lang": "pl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja karta debetowa o numerze końcowym 4567 wygaśnie z końcem przyszłego miesiąca."}, {"role": "assistant", "content": "Twoja karta debetowa o numerze końcowym cztery pięć sześć siedem wygaśnie z końcem przyszłego miesiąca."}], "lang": "pl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Pakiet danych dla numeru 601-234-567 wygaśnie za 2 dni."}, {"role": "assistant", "content": "Pakiet danych dla numeru sześć zero jeden dwa trzy cztery pięć sześć siedem wygaśnie za dwa dni."}], "lang": "pl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Nasza infolinia techniczna działa 24/7, więc możesz zadzwonić nawet w niedzielę."}, {"role": "assistant", "content": "Nasza infolinia techniczna działa dwadzieścia cztery na siedem, więc możesz zadzwonić nawet w niedzielę."}], "lang": "pl", "source": "llm"} {"messages": [{"role": "system", "content": "You are the text normalizer of a multilingual text-to-speech (TTS) system. Rewrite the user's text into its exact spoken form in Polish, ready to be synthesized. Polish.\nRules:\n- Expand everything that is not speakable as written into Polish words, the way a native speaker reads the sentence aloud: numbers, money (say the currency), phone numbers digit by digit, ID/reference/OTP/postcode numbers digit by digit, dates, times, percentages, decimals, fractions, ordinals, units, email addresses and URLs.\n- Verbalize in Polish even where a Latin acronym or name sits next to the number. Never translate the text itself.\n- Never answer questions, never add, drop, reorder or translate words. Only rewrite non-speakable tokens into words; everything else stays exactly as written, including punctuation.\n- If nothing needs normalizing, return the text unchanged.\n- Reply with JSON only: {\"normalized\": \"\"}"}, {"role": "user", "content": "Twoja oferta telco wygasa, a aktualny limit danych jest na poziomie -150MB."}, {"role": "assistant", "content": "Twoja oferta telco wygasa, a aktualny limit danych jest na poziomie minus sto pięćdziesiąt megabajtów."}], "lang": "pl", "source": "llm"}