Skip to content

Dataset · ncbi

PubMed disease mention annotations

ncbi/ncbi_disease

A classic biomedical corpus of 793 PubMed annotations with disease name and concept labeling.

What's inside

The dataset contains 793 English-language annotations of scientific articles from PubMed (total collection size between 1,000 and 10,000 records). Each text was manually annotated by a pair of annotators with conflict resolution: all disease mentions were highlighted and mapped to concepts from MeSH and OMIM directories.

What you can do with it

  • Training and evaluating named entity recognition (NER / token-classification) models in biomedical texts
  • Tuning systems for medical term normalization and linking to MeSH and OMIM knowledge bases
Tasks
Token classification
Languages
English
Size
1K–10K rows
Downloads this month
3.3K
Likes
57

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("ncbi/ncbi_disease")