Related Experiment Video
Updated: Dec 7, 2025
03:14
Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
Published on: December 6, 2024
882
Pre-trained language model augmented adversarial training network for Chinese clinical event detection
Zhi Chang Zhang1, Min Yu Zhang1, Tong Zhou1
1College of Computer Science and Engineering, Northwest Normal University, 967 Anning East Road, Lanzhou 730070, China.
Mathematical Biosciences and Engineering : MBE
|September 29, 2020
Summary
This study introduces a new AI model for Chinese clinical event detection (CED), improving semantic understanding and addressing class imbalance. The model achieved high precision, recall, and F1-score, especially for minority categories.
Area of Science:
- Medical Artificial Intelligence
- Natural Language Processing
- Computational Linguistics
Background:
- Clinical event detection (CED) is crucial in medical AI, but research on Chinese clinical narratives is limited due to scarce annotated data.
- Existing methods often overlook contextual information, hindering semantic understanding in Chinese medical texts.
Purpose of the Study:
- To develop and evaluate a novel approach for Chinese clinical event detection (CED).
- To enhance semantic understanding by integrating contextual information into character embeddings.
- To address the class imbalance problem common in medical datasets.
Main Methods:
- A novel encoder-decoder structure utilizing a pre-trained language model for Chinese CED.
- Integration of contextual representations into Chinese character embeddings to improve semantic understanding.
- Implementation of a punitive weight mechanism to manage class imbalance.
Main Results:
- The proposed model achieved superior performance compared to baseline models on a manually annotated Chinese corpus.
- Achieved a precision of 83.73%, recall of 86.56%, and F1-score of 85.12%.
- Demonstrated significant performance improvements on minority category samples, indicating robustness.
Conclusions:
- The novel encoder-decoder model effectively enhances Chinese clinical event detection by leveraging contextual information.
- The punitive weight strategy successfully addresses class imbalance, improving performance on underrepresented categories.
- This research contributes to advancing multilingual clinical event detection and medical AI for Chinese language data.
