mukulb's picture
Update README.md
4a6fb9d verified
metadata
license: cc-by-4.0
language:
  - en
pretty_name: MedQuAD - Medical QA Dataset
size_categories:
  - 10K<n<100K

MedQuAD - Medical Question Answering Dataset

Dataset Overview

MedQuAD (Medical Question Answering Dataset) is a collection of 16,407 medical question-answer pairs derived from 9 NIH websites. It covers 37 question types related to diseases, drugs, medical tests, and treatments. This dataset is useful for building medical question-answering models, retrieval-augmented generation (RAG) systems, and other NLP applications in the healthcare domain.

Dataset Structure

The dataset is stored in CSV format with the following columns:

Column Name Description
text A formatted string that includes both the question and answer.
query The question asked.
answers The corresponding answer.
topic_embeddings Placeholder for clustering or topic embeddings. (Default: -1)
group_name The medical category of the question-answer pair.

Example Entry

text: <HUMAN>: What is (are) Adult Acute Lymphoblastic Leukemia ?
      <ASSISTANT>: Key Points
                    - Adult acute lymphoblastic leukemia (ALL) is a type of cancer in which the bone marrow makes too many lymphocytes (a type of white blood cell).....
              
               
                    
query: What is (are) Adult Acute Lymphoblastic Leukemia ?
answers: Key Points
                    - Adult acute lymphoblastic leukemia (ALL) is a type of cancer in which the bone marrow makes too many lymphocytes (a type of white blood cell). ...
topic_embeddings: 1
group_name: 1_CancerGov_QA

Usage

You can load the dataset using Hugging Face Datasets Library:

from datasets import load_dataset

dataset = load_dataset("mukulb/clustered_MEDQUAD_dataset_with_groups")
print(dataset["train"][0])

Applications

  • Medical Question-Answering Systems
  • Retrieval-Augmented Generation (RAG) for Healthcare
  • Named Entity Recognition (NER) for Medical Concepts
  • Semantic Search and Information Retrieval

License

The dataset is available for research purposes only. Please refer to individual source websites for copyright information. https://github.com/abachaa/MedQuAD.git

Contribution

If you have any improvements or suggestions, feel free to open an issue or contribute via a pull request!


Maintained by: [Mukul Bedwa]
📧 Contact: [mukulbedwa.jrf01@iiitu.ac.in]
🚀 Hosted on Hugging Face: [your_hf_mukulbedwa/MedQuAD]