Elamur Research⌕
☰
home / papers / 2026 / IndicMedDialog: A Parallel Multi-Turn Medical Dialogue Dataset for Accessible Healthcare in Indic Languages

IndicMedDialog: A Parallel Multi-Turn Medical Dialogue Dataset for Accessible Healthcare in Indic Languages

Paper · 2026 · Shubham Kumar Nigam, Suparnojit Sarkar and Piyush Patel

Summary

Most medical dialogue systems answer in a single turn or are trained on template-built English data. This does not reflect how doctors narrow down a diagnosis through follow-up questions, and it leaves out speakers of Indic languages. The authors build IndicMedDialog, a parallel multi-turn consultation corpus in English and nine Indic languages. They enrich MDDial with synthetic dialogues generated by an LLM, translate everything with TranslateGemma, have native speakers check the output, and fix phonetic, lexical and spacing errors with a script-aware correction step. They then train IndicMedLM by applying LoRA to a quantized LLaMA-3.2-3B-Instruct. The model can optionally take patient background (age, gender, allergies) before it starts asking about symptoms. With semantic post-processing, Hindi, Marathi and Bengali reach solid accuracy, but Assamese, Tamil and Telugu fail almost completely. The authors trace these failures mainly to tokenizer and pretraining gaps in the base model. The released data, model and failure taxonomy give a starting point for affordable, safety-aware medical assistants in low-resource Indic languages.

Key points

  • Languages: Every dialogue in the corpus is aligned across English and nine Indic languages: Urdu, Telugu, Tamil, Punjabi, Marathi, Hindi, Gujarati, Bengali and Assamese.
  • Data: The 1,879 MDDial dialogues are combined with 1,101 synthetic ones from Llama-3.3-70B-Versatile, giving 2,980 dialogues covering 12 diseases and 118 symptoms. The MDDial test split has 237 dialogues.
  • Accuracy: IndicMedLM reaches 80.85% in English. After semantic post-processing it reaches 72.76% in Hindi, 68.51% in Marathi and 58.72% in Bengali. Strict label matching underrates Marathi by up to 55 percentage points.
  • Failure modes: The error analysis identifies five: Instruction Drift, Label Collapse, Cross-Domain Confusion, Tokenization/Truncation Failure and Paraphrase-over-Label Generation. Truncation problems cluster in Gurmukhi and Telugu scripts and do not appear in Devanagari.
  • Safety and recommendations: Precision for Traumatic Brain Injury is 0% in Assamese, Tamil and Telugu. The authors recommend LLM-as-a-Judge evaluation and propose fine-tuning on English with translation at inference time. Dataset, checkpoints and code are on GitHub.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓University of BirminghamBirmingham · United Kingdom27 works · 62,564 citations

Also on the byline: Madan Mohan Malaviya University of Technology · Heritage Institute of Technology.

 Fields, topics and concepts

📚 Shelves 8The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 worksMedicine · 275 worksHealth Informatics · 214 worksHealth Professions · 75 worksHealth Information Management · 56 works

🧭 Topics 1Research subjects and languages the work is about.Dialogue systems · 654 works

Cite

Cite as Shubham Kumar Nigam, Suparnojit Sarkar and Piyush Patel. 2026. IndicMedDialog: A Parallel Multi-Turn Medical Dialogue Dataset for Accessible Healthcare in Indic Languages. BioNLP 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.bionlp-1.84

@inproceedings{nigam2026indicmeddialog,
  title = {{IndicMedDialog: A Parallel Multi-Turn Medical Dialogue Dataset for Accessible Healthcare in Indic Languages}},
  author = {Shubham Kumar Nigam, Suparnojit Sarkar and Piyush Patel},
  booktitle = {BioNLP 2026},
  year = {2026},
  doi = {10.18653/v1/2026.bionlp-1.84},
  url = {https://doi.org/10.18653/v1/2026.bionlp-1.84}
}

References

18 references, 6 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Mara Finkelstein, Isaac Caswell, Tobias Domhan, Jan-Thorsten Peter, Juraj Juraska, Parker Riley, Daniel Deutsch, Geza Kovacs, Cole Dilanni, Colin Cherry, et al. 2026. Translategemma technical report. arXiv preprint arXiv:2601.09012. — [arXiv:2601.09012](https://arxiv.org/abs/2601.09012) find
  2. Lecheng Gong, Weimin Fang, Ting Yang, Dongjie Tao, Chunxiao Guo, Peng Wei, Bo Xie, Jinqun Guan, Zixiao Chen, Fang Shi, et al. 2026. Med-dialogrubrics: A comprehensive benchmark and evaluation framework for multi-turn medical consultations in large language models. arXiv preprint arXiv:2601.03023. — [arXiv:2601.03023](https://arxiv.org/abs/2601.03023) find
  3. Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Alex Vaughan, et al. 2024. The llama 3 herd of models. arXiv preprint arXiv:2407.21783.
  4. Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Liang Wang, Weizhu Chen, et al. 2022. Lora: Low-rank adaptation of large language models. Iclr, 1(2):3.
  5. Klaus Krippendorff. 2011. dorff’s alpha-reliability. find
  6. Computing krippen-Yunxiang Li, Zihan Li, Kai Zhang, Ruilong Dan, Steve Jiang, and You Zhang. 2023. Chatdoctor: A medical chat model fine-tuned on a large language model meta-ai (llama) using medical domain knowledge. Cureus, 15(6).
  7. Wenge Liu, Jianheng Tang, Yi Cheng, Wenjie Li, Yefeng Zheng, and Xiaodan Liang. 2022. Meddg: an entity-centric medical consultation dataset for entity-aware medical dialogue generation. In CCF International Conference on Natural Language Processing and Chinese Computing, pages 447–459. Springer. find
  8. Srija Macherla, Man Luo, Mihir Parmar, and Chitta Baral. 2023. Mddial: A multi-turn differential diagnosis dialogue dataset with reliability evaluation. arXiv preprint arXiv:2308.08147. — [arXiv:2308.08147](https://arxiv.org/abs/2308.08147) find
  9. Shubham Kumar Nigam, Suparnojit Sarkar, and Piyush Patel. 2026. Medaiddialog: A multilingual multi-turn medical dialogue dataset for accessible healthcare. arXiv preprint arXiv:2603.24132. — [arXiv:2603.24132](https://arxiv.org/abs/2603.24132) find
  10. Sara Pieri, Sahal Shaji Mullappilly, Fahad Shahbaz Khan, Rao Muhammad Anwer, Salman Khan, Timothy Baldwin, and Hisham Cholakkal. 2024. Bimedix: Bilingual medical mixture of experts llm. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 16984–17002. find
  11. Zihao Zhao, Eric Wallace, Shi Feng, Dan Klein, and Sameer Singh. 2021. Calibrate before use: Improving few-shot performance of language models. In International conference on machine learning, pages 12697–12706. Pmlr.
  12. Alejandro R. Salamanca, Diana Abagyan, Daniel D’souza, Ammar Khairi, David Mora, Saurabh Dash, Viraat Aryabumi, Sara Rajaee, Mehrnaz Mofakhami, Ananya Sahu, Thomas Euyang, Brittawnya Prince, Madeline Smith, Hangyu Lin, Acyr Locatelli, Sara Hooker, Tom Kocmi, Aidan Gomez, Ivan Zhang, Phil Blunsom, Ni… — [arXiv:2603.11510](https://arxiv.org/abs/2603.11510) find
  13. Vishal Vivek Saley, Goonjan Saha, Rocktim Jyoti Das, Dinesh Raghu, et al. 2024. Meditod: An english dialogue dataset for medical history taking with comprehensive annotations. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 16843–16877.
  14. Gemma Team, Thomas Mesnard, Cassidy Hardin, Robert Dadashi, Surya Bhupatiraju, Shreya Pathak, Laurent Sifre, Morgane Rivière, Mihir Sanjay Kale, Juliette Love, et al. 2024. Gemma: Open models based on gemini research and technology. arXiv preprint arXiv:2403.08295. find
  15. Tao Tu, Anil Palepu, Mike Schaekermann, Khaled Saab, Jan Freyberg, Ryutaro Tanno, Amy Wang, Brenna Li, Mohamed Amin, Nenad Tomasev, et al. 2024. Towards conversational diagnostic ai. arXiv preprint arXiv:2401.05654. — [arXiv:2401.05654](https://arxiv.org/abs/2401.05654) find
  16. Junda Wang, Zonghai Yao, Zhichao Yang, Huixue Zhou, Rumeng Li, Xun Wang, Yucheng Xu, and Hong Yu. 2024. Notechat: a dataset of synthetic patient-physician conversations conditioned on clinical notes. In Findings of the Association for Computational Linguistics: ACL 2024, pages 15183–15201. find
  17. Songhua Yang, Hanjie Zhao, Senbin Zhu, Guangyu Zhou, Hongfei Xu, Yuxiang Jia, and Hongying Zan. 2024. Zhongjing: Enhancing the chinese medical capabilities of large language model through expert feedback and real-world multi-turn dialogue. In Proceedings of the AAAI conference on artificial intellig… find
  18. Guangtao Zeng, Wenmian Yang, Zeqian Ju, Yue Yang, Sicheng Wang, Ruisi Zhang, Meng Zhou, Jiaqi Zeng, Xiangyu Dong, Ruoyu Zhang, et al. 2020. Meddialog: Large-scale medical dialogue datasets. In Proceedings of the 2020 conference on empirical methods in natural language processing (EMNLP), pages 9241–…

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/nigam-2026-indicmeddialog-parallel-multi-turn-medical/