An automatic approach for constructing a knowledge base of symptoms in Chinese

Area of Science:

• Medical Informatics
• Computational Linguistics
• Knowledge Representation

Background:

• Limited availability of Linked Open Data knowledge bases (KBs) in Chinese for life sciences.
• Necessity of Chinese KBs for automated processing and analysis of Chinese electronic medical records (EMRs).
• Critical need for a Chinese symptom KB as symptoms are foundational to clinical diagnosis.

Purpose of the Study:

• To develop and publish a public knowledge base of symptoms in Chinese.
• To supplement existing medical resources with a specialized Chinese symptom KB.
• To enable automated analysis of Chinese EMRs and support clinical diagnosis.

Main Methods:

• Designing a data schema referencing the Unified Medical Language System (UMLS).
• Extracting entities from healthcare and encyclopedia websites, training classifiers and Conditional Random Field (CRF) models for symptom extraction from EMRs.
• Fusing data from multiple sources, aligning entities and attributes, and linking the KB to UMLS for cross-lingual comparison.

Main Results:

• Publication of a KB with over 26,000 distinct Chinese symptoms, including traditional Chinese medicine symptoms and synonym pairs.
• Inclusion of related medical concepts like diseases and medicines, along with their relationships to symptoms.
• Successful linkage to UMLS, enabling analysis of differences between Chinese and English symptoms.

Conclusions:

• The developed Chinese symptom KB is a significant contribution to medical resources, addressing a critical gap.
• The KB facilitates the automated processing of Chinese EMRs and enhances clinical diagnosis.
• The KB is released as Linked Open Data, promoting accessibility and further research.