Construction And Application Of Knowledge Graph Of Commonly Used Chinese Materia Medica For Chinese Medicine Theory

With the gradual development of Chinese medicine society and the attention of the country,Chinese medicinal materials under the guidance of the theory of traditional Chinese medicine has achieved the role of treating diseases and regulating the body.Since each Chinese medicinal material has its own characteristics,it is very important to understand and master its characteristics for the compatible use of traditional Chinese medicine.Most existing search engines return inaccurate and cluttered information to users.In knowledge graph,knowledge is represented in a structured way,which contains rich semantic information and better correlation knowledge.The question answering method based on knowledge graph can provide users with intuitive and accurate answers by relying on its analysis and understanding of the semantic level of questions and precise query of the underlying knowledge graph.Therefore,this paper studies how to construct the knowledge graph of common Chinese medicinal materials for traditional Chinese medicine theory and design and implement the question answering method based on the knowledge graph:(1)To construct knowledge atlas of common Chinese medicinal materials for TCM theory.In this paper,the top-down construction method was used to design the conceptual model of knowledge atlas of Chinese medicinal materials,and then the extracted knowledge was filled into the knowledge base of the data layer according to the design of the pattern layer,and the knowledge storage was carried out to complete the construction of knowledge atlas of commonly used Chinese medicinal materials.Aiming at knowledge extraction,a BERT-Bi LSTM-CRF model for named entity recognition of Chinese medicinal materials was proposed,and P,R and F1 reached 95.31%,96.42% and 95.82%,respectively,on the test set.(2)A Tiny BERT-CNN model for intent classification of intent questions of Chinese medicinal materials was proposed to solve the problem of short text and slightly unbalanced data.Tiny BERT-CNN has obtained good performance on the task in this paper,which can guarantee better classification accuracy while consuming less resources.The values of P,R and F1 in the test set of intention questions of common Chinese medicinal materials were 96.4%,95.9% and 96.2%,respectively,according to Tiny BERT-CNN.(3)Based on the knowledge atlas of commonly used Chinese medicinal materials,this paper designed and completed the question and answer method of knowledge atlas of commonly used Chinese medicinal materials,and its main function was to provide users with knowledge and answer of Chinese medicinal materials.