queefwath/telegu-convserational-sft

Telugu Conversational Instruction-Response Dataset

1020 Telugu instruction→response pairs in everyday conversational register, built for SFT/distillation fine-tuning. Combines real spoken-Telugu data extracted from licensed corpora with LLM-generated pairs covering topics the corpora didn't reach.

Composition

Source	Count	Description
LLM-generated	543	Broad everyday topics: family, relationships, emotions, festivals, work, money, shopping, weather, hobbies, opinions, and more
AI4Bharat IndicVoices	359	Real spoken Telugu; instructions are AI4Bharat's own literal elicitation prompts (from their public prompt bank), matched to real recorded responses by content similarity
ARTPARK-IISc/Vaani	118	Real spoken Telugu image-description responses, sampled across 4 districts (Hyderabad, Guntur, Karimnagar, Chittoor)

Format

JSONL, one object per line:

{
  "id": "string",
  "instruction": "Telugu instruction/prompt text",
  "response": "Telugu response text",
  "category": "topic category",
  "register": "conversational",
  "source": "corpus:<dataset_name> | llm_generated",
  "license": "license string or null",
  "match_score": "float, corpus rows only — content-similarity score used to match instruction to response",
  "notes": "extraction/generation notes"
}

@inproceedings{DBLP:conf/acl/JavedNGJBMSAFPR24,
  author    = {Tahir Javed and Janki Nawale and Eldho Ittan George and Sakshi Joshi and
               Kaushal Santosh Bhogale and Deovrat Mehendale and Ishvindeer Virender Sethi and
               Aparna Ananthanarayanan and Hafsah Faquih and Pratiti Palit and Sneha Ravishankar and
               Saranya Sukumaran and Tripura Panchagnula and Sunjay Murali and Kunal Sharad Gandhi and
               Ambujavalli R and Manickam K. M. and C. Venkata Vaijayanthi and
               Krishnan Srinivasa Raghavan Karunganni and Pratyush Kumar and Mitesh M. Khapra},
  title     = {IndicVoices: Towards building an Inclusive Multilingual Speech Dataset for Indian Languages},
  booktitle = {ACL (Findings)},
  pages     = {10740--10782},
  publisher = {Association for Computational Linguistics},
  year      = {2024}
}

@misc{pulikodan2026vaanicapturinglanguagelandscape,
  title         = {VAANI: Capturing the language landscape for an inclusive digital India},
  author        = {Sujith Pulikodan and Abhayjeet Singh and Agneedh Basu and Nihar Desai and Pavan Kumar J and others},
  year          = {2026},
  eprint        = {2603.28714},
  archivePrefix = {arXiv},
  primaryClass  = {eess.AS},
  url           = {https://arxiv.org/abs/2603.28714}
}