Research And Implementation Of TCM Adverse Reaction Analysis System Based On Entity Recognition

In recent decades, the efficacy of traditional Chinese medicine (TCM) has been widely recognized at home and abroad. The emergence of TCM preparations makes the use of traditional Chinese medicine more convenient, which promotes the widespread application of traditional Chinese medicine. However, the reports of adverse reactions related to TCM has also increased. Therefore, it is of great significance for the public and medical workers to understand and master the adverse reactions of TCM. Most of the adverse reactions of TCM are included in the literature in the form of free text. Limited by the lack of basic research in Chinese natural language processing, there’s no relevant work in mining adverse reactions in Chinese literature.Under this background, we extract adverse reactions caused by TCM from various literature resources, and then construct TCM adverse reaction analysis system. The main work includes:1) Investigate the state of art of biomedical text mining, the related work and difficulties of Chinese medical text mining, and the related technologies of text mining.2) Analyze the data requirements and functional requirements of the system, design the overall system architecture and the realization of the function of each module. 3) Construct the TCM adverse drug reaction literature corpus, study the identification of adverse reaction entities in the literature. Three methods to improve the recognition effect of adverse reaction entities are proposed:method based on CRF and domain knowledge; method based on text categorization and CRF; method based on text categorization, CRF and domain knowledge.4) Implement TCM adverse reaction analysis system based on the adverse reaction knowledge extracted form literature resources using methods mentioned above. The system provides services such as TCM adverse reaction retrieval and display, similar literature recommendation, real-time adverse reaction entity recognition.