Annotated Drug Reviews for Aspect Term Extraction and Polarity Detection

ATEdrug is a dataset of patient-generated drug reviews for three medical conditions: Depression, Arthritis, and Birth control. The dataset was annotated for Aspect Term Extraction and Polarity Detection using an expert-driven rule-based approach with minimal human intervention. It was created by Gunjan Ansari and last updated on 2026-04-09.

Use Cases

• Train transformer models like BERT for Aspect Term Extraction based on annotated drug reviews.
• Develop models for Polarity Detection of medical aspects mentioned in patient opinions.
• Benchmark automated annotation schemes against human evaluation for clinical text.
• Facilitate research in pharmacovigilance by providing labeled data for adverse reaction monitoring.

Strengths

• Dataset is publicly available under a CC-BY-4.0 license.
• Annotation was validated through human annotator agreement.
• Includes experimental code and data on a linked GitHub repository.

Limitations

• Dataset size is 5.5 KB, indicating a very limited scope.
• Row count and column-level documentation are unknown.
• Focus is limited to three specific medical conditions.

Provenance

Source

figshare, author Gunjan Ansari.

Collection Method

Drug reviews annotated via an expert-driven rule-based approach (ATEdrug) with human validation.

Time Range

null

Freshness

Last updated 2026-04-09 17:35:07.

Geography

null