Related Experiment Video
Updated: Oct 11, 2025
08:15
Author Spotlight: Network Pharmacology and Molecular Docking to Decipher the Action of Jiawei Shengjiang San Against Diabetic Kidney Disease
Published on: May 10, 2024
695
Chinese-Named Entity Recognition From Adverse Drug Event Records: Radical Embedding-Combined Dynamic Embedding-Based
Hong Wu1, Jiatong Ji2, Haimei Tian3
1School of Science, China Pharmaceutical University, Nanjing, China.
JMIR Medical Informatics
|December 2, 2021
Summary
This study introduces BBC-Radical, an automated tool for identifying adverse drug reaction (ADR) information in Chinese reports. The model significantly improves upon manual methods for enhanced drug safety evaluation.
Area of Science:
- Pharmacovigilance and Drug Safety
- Natural Language Processing in Healthcare
- Computational Linguistics
Background:
- Adverse drug events (ADEs) are increasing with drug variety.
- Electronic medical records and ADR reports are key sources for ADR information.
- Automating the extraction of latent ADR information is crucial for drug safety.
Purpose of the Study:
- To describe a method for identifying ADR-related information from Chinese ADE reports.
- To develop and evaluate an automated tool for ADR information extraction.
Main Methods:
- Developed BBC-Radical, a model combining BERT, bi-LSTM, and CRF.
- Utilized token and radical features of Chinese characters for Named Entity Recognition (NER).
- Trained and tested the model on 24,890 ADR reports and compared performance against a manual method.
Main Results:
- The BBC-Radical model achieved high NER performance (precision 87.2%, recall 85.7%, F1 score 86.4%).
- The automated model significantly outperformed the manual method in entity recognition.
- The underlying NER model demonstrated excellent performance (precision 96.4%, recall 96.0%, F1 score 96.2%).
Conclusions:
- The BBC-Radical model is effective for extracting ADR information from reports.
- This method is significant for improving ADR report quality and postmarketing drug safety.
- Automated extraction enhances pharmacovigilance and drug safety reevaluation.
