| TCM（Traditional Chinese Medicine）has a long history,low treatment cost and less side effects.Especially in the treatment of some specific diseases TCM often has unexpected effects.In the era of information explosion,there are more and more literatures in the field of TCM.How to use these data through text mining and other technologies is not only a challenge,but also an opportunity.At present,there are few open-source clinical knowledge graphs of TCM,which is very unfavorable for the development and further research of TCM.In order to solve this problem,this study used unstructured Chinese text data to construct the clinical knowledge graph of TCM.Firstly,a large number of TCM clinical related entities were collected and integrated manually,and the information is used as external dictionary.Then,a two-level TCM clinical corpus is constructed based on a large number of Chinese literature reports.The first-level corpus is mainly composed of sentences that appear in pairs among the five categories of entities:herb,prescription,TCM disease,TCM symptom and TCM syndrome.This corpus can be used to search for possible unlisted TCM clinical entities through juxtaposition,and to evaluate the correlation between TCM clinical entities through co-occurrence word frequency analysis.The co-occurrence sentences are automatically labeled with BIOSE according to the TCM clinical entities in the existing external dictionary to form a second-level corpus.The second-level corpus is mainly used for entity recognition using HMM,CRF,BILSTM and BILSTM-CRF models.It can also be directly used in other natural language processing tasks.Finally,this paper summarizes the correlation results between TCM clinical entities and describes the two-way frequency relationship between entities as an information flow relationship through the concept of information cohesiveness degree proposed in this paper.According to the degree of information cohesiveness,the data is processed into triples,and the graph database is established on the local server through neo4j,and the triples are imported into the graph database.Users can easily use graph database query language Cypher to query related data in TCM clinical knowledge graph and the query results will be displayed in the form of graphics.Subsequently,this article also proposes an application that correlates TCM symptoms with Western medicine symptoms,and analyzes TCM symptoms from the perspective of molecular mechanisms.The research results can assist the clinical diagnosis and treatment of traditional Chinese medicine to a certain extent,and it has also taken a key step in breaking the barriers between Chinese and Western modern medicine. |
