Research And Implementation Of Automatic Question Answering System Based On Medical Knowledge Graph

The advent of the cloud era has greatly changed people’s lives.With the vigorous development of information technology and artificial intelligence,people’s medical and health needs are also rising.Therefore,how to build a simple,practical and popular disease diagnosis and treatment system for the benefit of people’s health has become the key to the research and application of smart medicine.Traditional medical search engine technology indexes related web pages by keywords,which is inefficient and needs artificial intelligence to change the status quo.As a branch of artificial intelligence,knowledge graph started relatively late.However,as the application direction of knowledge graph,the development of question answering system is not perfect,which makes the intelligent medical research complex and difficult to carry out.Thus,this paper proposes to collect and sort out a large number of medical datas,constructs a large-scale maintainable medical knowledge graph,studys and summarizes the key technologies of knowledge graph and question answering system,integrates and optimizes them,designs and implements an auxiliary medical product,namely the automatic question answering system based on medical knowledge graph,as the first line of defense to protect people’s healthy livings.The research work of this paper mainly includes:(1)In view of the complex and scattered medical information data,this paper focuses on how to obtain,process,fuse and store multi-source data sets,and integrate and optimize the massive medical data sets to build a reliable medical knowledge graph for the system.The alternating combination method is designed to define entities,relationships and attributes in advance,and the ontology idea is introduced into the construction of data schema,which makes the construction of knowledge graph more convenient.This paper proposes a storage scheme based on hybrid database to provide efficient storage of knowledge graph and describe the whole process of building knowledge graph.(2)Aiming at people’s complex and diverse medical problems,this paper puts forward the solution of question answering system suitable for medical service.Firstly,the workflow of the system is designed,including the construction scheme of problem analysis and answer query.Then,the problem model and analysis method based on template matching are designed,and a semantic similarity calculation method composed of editing distance,character overlap coefficient and word vector is proposed,and the improved scheme is designed.Experiments show that the method can effectively extract disease and symptom entities with similar foreign language meanings in dictionaries.Considering the problem of database query target,a multi classification model method for intention recognition is designed.In the test experiment,the best F1 value of the multi classification model reaches 0.95,which shows that the multi classifier has a good effect on solving the query intention of user input information.(3)Aiming at the problem that the current search engine technology for medical question answering system is backward,the medical knowledge graph and question answering system technology are integrated and optimized to realize the construction and implementation of intelligent medical question answering system.Finally,the response time and performance of the system are tested and analyzed,which proves that the automatic question answering system researched and implemented in this paper is applicable and feasible,and has certain application value.