MedNLPCombined

Dataset Description

MedNLPCombined is a collected repository of medical Natural Language Processing (NLP) datasets, primarily focused on Chemical-Disease Relations (CDR), Toxicogenomics, and Gene Interactions. This repository is designed to facilitate research in Named Entity Recognition (NER) and Relation Extraction (RE) within the biomedical domain.

The repository currently includes six major components:

1. BioCreative V CDR (BC5CDR) Task Corpus
2. Comparative Toxicogenomics Database (CTD) Derived Data
3. ChemDisGene Dataset
4. MedQA Dataset
5. PubMedQA Dataset
6. BioASQ Dataset

Repository Structure

The dataset is organized into the following directories:

1. bc5cdr/

Contains the BioCreative V CDR corpus resources.

• data/: The core dataset files.
  • benchmark/: Evaluation/test sets.
  • training/: Training and development sets.
• related_documents/: Documentation and guidelines for the BC5CDR task.
  • BC5CDR.corpus.pdf
  • BC5CDR.overview.pdf
  • bc5_CDR_data_guidelines.pdf

2. CTD/

Contains data derived from the Comparative Toxicogenomics Database.

• CTD.zip: A compressed archive likely containing processed abstracts, relationships, and entity annotations derived from CTD.
  • Note: You may need to unzip this file to access the raw .tsv or .txt files.

3. ChemDisGene/

Contains the ChemDisGene dataset for distant supervision of biomedical relationships.

• data/: The core dataset files.
• related_documents/: Documentation and guidelines.
  • AnnotationGuidelines.pdf

4. medqa/

Contains the MedQA dataset.

• A large-scale open-domain multiple-choice dataset for medical problems, collected from professional exams.

5. pubmedqa/

Contains the PubMedQA dataset.

• A biomedical research question answering dataset requiring reasoning over PubMed abstracts.

6. bioasq/

Contains the BioASQ dataset.

• A benchmark dataset for large-scale biomedical semantic indexing and question answering.

Dataset Details

BioCreative V CDR (BC5CDR)

The BC5CDR corpus consists of 1,500 PubMed articles with annotated chemical and disease entities, as well as their chemical-induced disease (CID) relations. It was created for the BioCreative V challenge.

• Entities: Chemicals, Diseases
• Relations: Chemical-Induced Disease (CID)

Comparative Toxicogenomics Database (CTD)

The CTD data provides manually curated information about chemical-gene/protein interactions, chemical-disease and gene-disease relationships. This component of the repository is likely a snapshot or a derived subset focusing on specific interaction types (e.g., chemical-disease-gene networks).

ChemDisGene

ChemDisGene is a large-scale, distant-supervision dataset for extracting biomedical relationships between chemicals, diseases, and genes. It provides a valuable resource for training models on a broader range of biomedical interactions. The dataset contains approximately 80,000 biomedical research abstracts annotated with mentions of chemical, disease, and gene/gene-product entities, along with their pairwise relationships.

• Paper: A Distant Supervision Corpus for Extracting Biomedical Relationships Between Chemicals, Diseases and Genes
• GitHub: chanzuckerberg/ChemDisGene
• Hugging Face: bigbio/chem_dis_gene

MedQA

MedQA is a large-scale open-domain question answering dataset derived from professional medical board exams in the US, Mainland China, and Taiwan. It evaluates models on professional medical knowledge and clinical decision-making through multiple-choice questions.

• Paper: What Disease does this Patient Have? A Large-scale Open Domain Question Answering Dataset from Medical Exams
• GitHub: jind11/MedQA
• Hugging Face: bigbio/med_qa

PubMedQA

PubMedQA is a biomedical question answering dataset designed to answer research questions with yes/no/maybe using the corresponding abstracts. It requires reasoning over quantitative content and scientific texts.

• Paper: PubMedQA: A Dataset for Biomedical Research Question Answering
• GitHub: pubmedqa/pubmedqa
• Hugging Face: bigbio/pubmed_qa

BioASQ

BioASQ is a large-scale biomedical semantic indexing and question answering benchmark dataset. It provides various QA tasks challenging systems with realistic information needs from biomedical experts.

• Paper: An overview of the BIOASQ large-scale biomedical semantic indexing and question answering competition
• GitHub: BioASQ Organization
• Hugging Face: bigbio/bioasq_task_b

Usage

To use this dataset in your Python project:

from datasets import load_dataset

# Example loading (if configured with a loading script, otherwise access files directly)
# dataset = load_dataset("username/MedNLPCombined")

For local usage, clone the repository:

git clone https://huggingface.co/username/MedNLPCombined

Citation

If you use these datasets, please cite the original authors:

BC5CDR:

@article{li2016biocreative,
  title={BioCreative V CDR task corpus: a resource for chemical disease relation extraction},
  author={Li, Jiao and Sun, Yueping and Johnson, Robin J and Sciaky, Daniela and Wei, Chih-Hsuan and Leaman, Robert and Davis, Allan Peter and Mattingly, Carolyn J and Lu, Zhiyong},
  journal={Database},
  volume={2016},
  year={2016},
  publisher={Oxford Academic}
}

CTD: Please refer to the CTD citation policy.

ChemDisGene:

@inproceedings{zhang-etal-2022-distant,
    title = "A Distant Supervision Corpus for Extracting Biomedical Relationships Between Chemicals, Diseases and Genes",
    author = "Zhang, Dongxu  and
      Mohan, Sunil  and
      Torkar, Michaela  and
      McCallum, Andrew",
    booktitle = "Proceedings of the Thirteenth Language Resources and Evaluation Conference",
    month = jun,
    year = "2022",
    address = "Marseille, France",
    publisher = "European Language Resources Association",
    url = "https://aclanthology.org/2022.lrec-1.666",
    pages = "6205--6214",
}

MedQA:

@article{jin2020disease,
  title={What disease does this patient have? a large-scale open domain question answering dataset from medical exams},
  author={Jin, Di and Pan, Eileen and Oufattole, Nassim and Weng, Wei-Hung and Fang, Hanyi and Szolovits, Peter},
  journal={arXiv preprint arXiv:2009.13081},
  year={2020}
}

PubMedQA:

@inproceedings{jin2019pubmedqa,
  title={PubMedQA: A Dataset for Biomedical Research Question Answering},
  author={Jin, Qiao and Dhingra, Bhuwan and Liu, Zhengping and Cohen, William and Lu, Xinghua},
  booktitle={Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP)},
  pages={2567--2577},
  year={2019}
}

BioASQ:

@article{tsatsaronis2015overview,
  title={An overview of the BIOASQ large-scale biomedical semantic indexing and question answering competition},
  author={Tsatsaronis, George and Balikas, Georgios and Malakasiotis, Prodromos and Partalas, Ioannis and Zschunke, Matthias and Alvers, Michael R and Weissenborn, Dirk and Krithara, Anastasia and Petasis, Pethes and Polychronopoulos, Dimitris and others},
  journal={BMC bioinformatics},
  volume={16},
  number={1},
  pages={1--28},
  year={2015},
  publisher={BioMed Central}
}

License

This repository is licensed under Apache 2.0. Please also adhere to the specific license agreements of the original datasets (BC5CDR, CTD, ChemDisGene, MedQA, PubMedQA, and BioASQ) if applicable.