MedNLP-DataFactory: MedQA-Synth-Raw

📑 Paper | 🌐 Project Page | 💾 Released Resources | 📦 Repo

This is the raw medical QA dataset used in our MedNLP-DataFactory project. The data was carefully curated and filtered from the original MedMCQA dataset developed by Pal et al.

Data Source

The raw data in this dataset is directly adopted from the MedMCQA dataset, a large-scale multiple-choice question answering dataset derived from medical entrance exams. We performed additional filtering to retain only clinically relevant questions suitable for reasoning chain synthesis.

The data format for each line in the medqa_filtered.jsonl is as follows:

{
  "question": "<the medical question>",
  "options": "<answer options A, B, C, D>",
  "correct_answer": "<the correct option>",
  "explanation": "<brief explanation if available>",
  "subject": "<medical subject area>",
  "source": "MedMCQA"
}

Citation

If you use this dataset, please cite the original MedMCQA paper:

@inproceedings{pal2022medmcqa,
  title={MedMCQA: A Large-scale Multi-Subject Multi-Choice Dataset for Medical domain Question Answering},
  author={Pal, Ankit and Umapathi, Logesh Kumar and Sankarasubbu, Malaikannan},
  booktitle={Conference on Health, Inference, and Learning},
  year={2022}
}

License

The license for this dataset is apache-2.0.