| In order to further expand and extend the ability of data expression,this paper puts forward the definition and expression method of relation-strength mechanism of knowledge graph,and puts forward the concepts of relation-strength factor and relation-time sequence factor,aiming at improving the ability of data expression between entities and providing more data support for Knowledge Computing.The extraction process of relation-strength factor based on dependency parsing is designed and applied in medical field.Among the many standards to evaluate social progress,medical and health knowledge is a very important one.However,there are many thorny problems in the field of health care: high medical expenditure,high cost of medical infrastructure replacement,high technical requirements of medical practitioners and so on.The emergence of knowledge graph is to promote the rapid development of Medical Knowledge QA system.This paper constructs a Chinese medical knowledge graph,and implements a medical knowledge QA system on the basis of the knowledge base.Through many related technologies of NLP,it establishes a relevant language model,extracts medical knowledge from the semi-structured data of Internet sites,realize medical knowledge retrieval and question answering.In this paper,we collect and organize the medical knowledge data from Internet sites through Python script,and transform it into structured knowledge data after data pre-processing.According to the characteristics of medical knowledge,the ontology data specification of knowledge graph is designed,and the knowledge data is imported based on neo4 j graphic database,and finally the medical domain knowledge graph is constructed;by analyzing the feature space of users’ natural language questions,the problem intention is identified through Text CNN model,the AC-automaton is established,and the basic types of questions are identified,and then the question rewriter is used The question is translated into cypher query,the knowledge graph is retrieved and the answer is optimized,the retrieval and question answering of medical knowledge data are completed,the multiple question types are supported,and the synchronous recognition of multiple question types is realized;the practical application of relation-strength mechanism in the medical knowledge QA system is realized,and the front-end of human-computer interaction is completed by the FLASK web container,which is convenient for users to use and operate. |
