---
license: apache-2.0
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
dataset_info:
  features:
  - name: id
    dtype: int64
  - name: title
    dtype: string
  - name: authors
    dtype: string
  - name: journal
    dtype: string
  - name: content
    dtype: string
  - name: source_url
    dtype: string
  - name: publication_types
    sequence: string
  - name: pubmed_id
    dtype: string
  splits:
  - name: train
    num_bytes: 1001771206
    num_examples: 471336
  download_size: 525136760
  dataset_size: 1001771206
task_categories:
- question-answering
language:
- en
tags:
- medical
- therapeutics
pretty_name: Pubmed-Medline
---
# PubMed Medline
## Overview

This dataset contains metadata and abstracts extracted from biomedical research publications indexed in **PubMed** (Date Accessed: 05/02/2025). The dataset is filtered to include only MEDLINE-indexed articles in the English language. Each record represents a single research publication with key metadata fields that provide insights into the article's content, authorship, and publication details.

## Dataset Structure

Each record in the dataset represents a single PubMed article and contains the following fields:

* **id**: The unique identifier for the article (integer PubMed ID)
* **pubmed_id**: The PubMed ID as a string format (for compatibility with external systems)
* **title**: The full title of the research article
* **authors**: Comma-separated list of author names
* **journal**: Publication details including journal name, volume, issue, and publication date
* **content**: The article abstract text
* **source_url**: Direct link to the article on PubMed (format: https://pubmed.ncbi.nlm.nih.gov/{id}/)
* **publication_types**: Array of publication types (e.g., "Journal Article", "Review", "Clinical Trial")

## Acknowledgements

We acknowledge the National Library of Medicine (NLM) for maintaining PubMed and making biomedical literature metadata accessible for research purposes.