Research And Application Of Traditional Chinese Medicine Knowledge Relation Discovery

Today it has entered the era of big data. How to discover and organize knowledge to build knowledge graph has received great attention in various of fields. The construction of domain knowledge graph confronts many challenges which includes research of specific problems and technology.As the cultural treasure of China, Traditional Chinese medicine has produced rich data in the fusion of modern science and discipline, which builds the foundation for the construction of domain knowledge graph. Under the background of China Knowledge Center for Engineering Science and Technology, this thesis focus the method of knowledge discovery and the construction of Chinese traditional medicine knowledge graph lexicon system. It includes following content.1.On the basis of study of Traditional Chinese Medicine Thesauri, the dissertation proposes a Bayes Net-based method to discover the hyponymy relations. With this method, the number of entries of Traditional Chinese Medicine Thesauri has been expanded from 13605 to 146258. The experiment shows that precision rate and recall rate are both above 90%, which proves the method is effective.2.With the data of library and wiki, the dissertation proposes three kinds of relation types. They are Cure Relationship, Manifestation relationship and Composition Relationship. For Cure Relationship and Manifestation relationship, I propose a method which is based on SVM and word embedding to discover these two types of relation. for Composition Relation,I use a string matching-based method to discover this type of relation.3. A Traditional Chinese Medicine Lexicon System is built under construction of the above methods. The system provides service such as word search, relation search, word segmentation computation, classification computation and word similarity computation.