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

The Internet has become the first choice for people to obtain information,but the traditional retrieval methods often return some disorderly web links,which need to be filtered manually.At the same time,with the influx of a large number of professional fields of information,the challenge of identifying effective information is also gradually increasing.Interactive question answering(IQA)analyzes the user’s questions and returns accurate and intuitive answers through question parsing.This more efficient and intelligent query method meets the needs of people in modern society for accurate and efficient access to information.Taking the medical field as the application scenario,this paper proposes the RBLGP model to identify medical nested entities and the RBLSA model to identify the intention of questions.On this basis,the construction of medical knowledge graph and the construction of question answering system are completed.The main contributions of this paper are as follows:(1)Construction of Knowledge Graphs in the Medical Domain.Aiming at the problem of nested entities in entity extraction tasks in the medical field,an entity recognition model based on Ro BERTa-wwm+Bi LSTM+Global Pointer(RBLGP)is proposed.The Ro BERTa-wwm pre-training model and Bi LSTM network are introduced to extract text features,and the Global Pointer is used to complete the labeling task to improve the accuracy of entity recognition,and the efficiency of recognition is also significantly improved.On this basis,a model named RBLGPL for joint extraction of entities and relations from unstructured data is designed.The model is based on GPLinker to solve the nesting problem of joint extraction of entities and relations.Finally,the construction of medical knowledge graph was completed by storing the multi-source data of knowledge fusion in the Neo4 j graph database.(2)Research on Algorithms of Question Answering System.Aiming at the problem that the traditional question intention classification method is not accurate enough,this thesis proposed a model RBLSA based on Ro BERTa-wwm+Bi LSTM+Self-Attention.The model used Ro BERTa-wwm as word vector embedding,and Bi LSTM network learned the semantic information between questions.The fusion between sentences is realized by calculating the Self-Attention weight,which enhances the learning ability of the model and obtains better classification results.Based on the result of question parsing,the answer query method is designed to return the content that conforms to the user’s intention.(3)Implementation of Question Answering System Based on Knowledge Graph in Medical domain.Through the requirement analysis of the medical question answering system,based on the improved recognition algorithm and question parsing algorithm in this paper,the medical knowledge graph is used as the knowledge base of the question answering system,and the medical knowledge question answering system is designed and implemented.The system includes medical question answering,query visualization and recognition functions,and the performance of the system is tested.Finally,the question answering system has good interaction with users,meets the needs of multiple parties,and has high practicability.