Related Experiment Video
Updated: Sep 3, 2025
05:56
Objectification of Tongue Diagnosis in Traditional Medicine, Data Analysis, and Study Application
Published on: April 14, 2023
2.6K
Multi-level semantic fusion network for Chinese medical named entity recognition
Jintong Shi1, Mengxuan Sun1, Zhengya Sun1
1University of Chinese Academy of Sciences, Beijing, 100049, China; Institute of Automation, Chinese Academy of Sciences, Beijing, 100190, China.
Journal of Biomedical Informatics
|July 25, 2022
Summary
This study introduces a novel multi-level semantic fusion network for Chinese medical named entity recognition (MNER). The model effectively integrates hierarchical semantics, improving entity extraction from complex medical texts.
Area of Science:
- Natural Language Processing
- Medical Informatics
- Computational Linguistics
Background:
- Medical Named Entity Recognition (MNER) is crucial for analyzing electronic health records.
- Existing MNER methods often overlook hierarchical semantic information.
- Chinese medical texts present unique challenges due to homophones and pictophonetic characters.
Purpose of the Study:
- To develop an advanced MNER model for Chinese medical texts.
- To leverage multi-level semantic information, including morphology, character, word, and syntax.
- To enhance the understanding of unstructured medical data.
Main Methods:
- Proposed a multi-level semantic fusion network.
- Fused morphological (radical), character (pinyin, dictionary), and word (dictionary) semantics using BiLSTM.
- Employed a graph neural network to incorporate syntactic (word dependency) semantics.
Main Results:
- Demonstrated the effectiveness of the proposed model on two public datasets.
- Achieved robust performance in real-world scenarios.
- The multi-level fusion approach significantly improved MNER accuracy.
Conclusions:
- The multi-level semantic fusion network offers a superior approach to Chinese MNER.
- Integrating diverse semantic levels enhances the model's ability to interpret complex medical texts.
- This method holds promise for advancing medical text analysis and information extraction.
