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

The automatic question answering system has gradually become the new trend of the natural interaction between human and machine.Knowledge graph enables data to be better organized and understood in a form that is close to human cognition,and can play a key role in precision question answering services.As the service construction in the medical and health field is still not perfect,China has launched the policy of intelligent medical construction.Based on the project demand of intelligent medical service of National Health Commission of China,this paper selects the application of question answering system in the medical vertical field to start the research.However,there is a lack of public clinical knowledge graph in the Chinese field at present,and it is difficult to construct high-quality knowledge graph in the medical field.Question answering system based on knowledge graph needs strong natural question comprehension ability,while medical field lacks training corpus of question answering based on knowledge base,the question answering task based on deep learning is full of challenges in practice.In view of the above problems,this paper proposes a solution of knowledge graph construction and automatic question answering system based on knowledge graph centering on the clinical medical field,and completes the design and implementation of the medical auxiliary service platform.The main research contents include the following three points:(1)Research on how to construct the knowledge graph of clinical medical field.Mainly includes the knowledge extraction,knowledge fusion and knowledge storage of three parts,among them,the automatic extraction of knowledge is realized by using an entity relation extraction model based on Bi-LSTM-CRF network and jointly tagging strategies,the fusion of knowledge is realized by using an entity alignment method based on local medical entity property and relation similarity,the storage of knowledge is realized by using a mixed way with Neo4j and MongoDB.(2)Research how to realize the automatic question answering system based on medical knowledge graph.At first,the semantic analysis of questions is realized by using question entity recognition based on the Lattice-LSTM-CRP network and relation/prope-ty mapping based on CNN with char-word embedding.Secondly,the answer retrieval of knowledge base is accomplished by constructing semantic query logic rule convertor.(3)Build an auxiliary medical service platform.Based on the above research,the medical consultation service with knowledge base was realized,and the graph relationship visualization service was realized based on the force-directed graph with D3.j s.Finally,the service platform integrating medical professional knowledge retrieval and registration was completed,and the WeChat medical assistant was realized.Through the study of the above content,the purpose of this paper is to solve the difficulty in the process of actual construction in Chinese clinical medicine knowledge graph,and through the deep learning method improve questions semantic understanding ability of medical knowledge graph automatic question answering system,eventually to build a precision medical auxiliary service platform that can fulfill user's real medical demands.