hat_tts / README.md
hungshinlee's picture
Update Dataset Card (concise subset statistics)
df4593c verified
|
Raw History Blame Contribute Delete
4.95 kB
metadata
dataset_info:
  - config_name: hak-x-hailu
    features:
      - name: id
        dtype: string
      - name: audio
        dtype: audio
      - name: duration
        dtype: float64
      - name: hanzi
        dtype: string
      - name: pinyin
        dtype: string
      - name: ipa
        dtype: string
      - name: mandarin
        dtype: string
      - name: language
        dtype: string
      - name: lang_code
        dtype: string
      - name: speaker
        dtype: string
      - name: snr
        dtype: float64
      - name: bandwidth_hz
        dtype: int64
      - name: real_upper_frequency
        dtype: int64
      - name: bandwidth_category
        dtype: string
      - name: is_upsampled
        dtype: bool
      - name: dnsmos_ovrl
        dtype: float64
      - name: dnsmos_sig
        dtype: float64
      - name: dnsmos_bak
        dtype: float64
      - name: dnsmos_p808
        dtype: float64
      - name: emotion_label
        dtype: string
      - name: emotion_score
        dtype: float64
      - name: emotion_scores
        struct:
          - name: angry
            dtype: float64
          - name: disgusted
            dtype: float64
          - name: fearful
            dtype: float64
          - name: happy
            dtype: float64
          - name: neutral
            dtype: float64
          - name: other
            dtype: float64
          - name: sad
            dtype: float64
          - name: surprised
            dtype: float64
          - name: unknown
            dtype: float64
      - name: age
        dtype: float64
      - name: age_group
        dtype: string
      - name: age_group_code
        dtype: string
      - name: gender_label
        dtype: string
      - name: gender_score
        dtype: float64
      - name: gender_probabilities
        struct:
          - name: female
            dtype: float64
          - name: male
            dtype: float64
          - name: child
            dtype: float64
      - name: num_speakers
        dtype: int64
      - name: is_single_speaker
        dtype: bool
      - name: speaker_confidence
        dtype: float64
      - name: spk_embedding
        list: float64
      - name: spk_consistency
        dtype: float64
      - name: speaking_rate
        dtype: float64
      - name: loudness_lufs
        dtype: float64
      - name: suggested_gain_db
        dtype: float64
      - name: quality_tier
        dtype: string
      - name: quality_tier_label
        dtype: string
      - name: split
        dtype: string
    splits:
      - name: train
        num_bytes: 8623835207
        num_examples: 24081
    download_size: 9572599253
    dataset_size: 8623835207
  - config_name: hak-x-sixian
    features:
      - name: id
        dtype: string
      - name: audio
        dtype: audio
      - name: duration
        dtype: float64
      - name: hanzi
        dtype: string
      - name: pinyin
        dtype: string
      - name: ipa
        dtype: string
      - name: mandarin
        dtype: string
      - name: language
        dtype: string
      - name: lang_code
        dtype: string
      - name: speaker
        dtype: string
      - name: snr
        dtype: float64
      - name: bandwidth_hz
        dtype: int64
      - name: real_upper_frequency
        dtype: int64
      - name: bandwidth_category
        dtype: string
      - name: is_upsampled
        dtype: bool
      - name: dnsmos_ovrl
        dtype: float64
      - name: dnsmos_sig
        dtype: float64
      - name: dnsmos_bak
        dtype: float64
      - name: dnsmos_p808
        dtype: float64
      - name: emotion_label
        dtype: string
      - name: emotion_score
        dtype: float64
      - name: emotion_scores
        struct:
          - name: angry
            dtype: float64
          - name: disgusted
            dtype: float64
          - name: fearful
            dtype: float64
          - name: happy
            dtype: float64
          - name: neutral
            dtype: float64
          - name: other
            dtype: float64
          - name: sad
            dtype: float64
          - name: surprised
            dtype: float64
          - name: unknown
            dtype: float64
      - name: age
        dtype: float64
      - name: age_group
        dtype: string
      - name: age_group_code
        dtype: string
      - name: gender_label
        dtype: string
      - name: gender_score
        dtype: float64
      - name: gender_probabilities
        struct:
          - name: female
            dtype: float64
          - name: male
            dtype: float64
          - name: child
            dtype: float64
      - name: num_speakers
        dtype: int64
      - name: is_single_speaker
        dtype: bool
      - name: speaker_confidence
        dtype: float64
      - name: spk_embedding
        list: float64
      - name: spk_consistency
        dtype: float64
      - name: speaking_rate
        dtype: float64
      - name: loudness_lufs
        dtype: float64
      - name: suggested_gain_db
        dtype: float64
      - name: quality_tier
        dtype: string
      - name: quality_tier_label
        dtype: string
      - name: split
        dtype: string
    splits:
      - name: train
        num_bytes: 19384647413
        num_examples: 26315
    download_size: 10132187591
    dataset_size: 19384647413
configs:
  - config_name: hak-x-hailu
    data_files:
      - split: train
        path: hak-x-hailu/train-*
  - config_name: hak-x-sixian
    data_files:
      - split: train
        path: hak-x-sixian/train-*

Dataset Information

  • Corpus: hat_tts
  • Status: Public (Gated)

Subset Stats

Subset Train (Samples / Hours) Eval (Samples / Hours)
hak-x-hailu 24,081 / 55.10 0 / 0.00
hak-x-sixian 26,315 / 58.29 0 / 0.00
Total 50,396 / 113.39 0 / 0.00