Research And Implementation Of A Knowledge Graph-based Question And Answer System For Liver Diseases

According to statistics,the number of chronic liver disease patients in China exceeded 247 million in 2020.Liver disease is seriously eroding the physical and mental health of the nation.However,too many patients often lack a full understanding of liver disease,which leads to the delay of the best time for treatment and even aggravation of the disease.The rapid development of the Internet industry has brought us into an era of information explosion.How to efficiently and conveniently retrieve the information we need in the jumble of Internet information has become a big problem for current researchers.The development of knowledge graph and question answering system provides convenience for intelligent query.In this paper,the task of named entity recognition is studied based on the deep learning model,so as to enhance the question parsing ability of the system.Combined with the construction of knowledge graph,a question and answer system about liver diseases is built to provide users with scientific and accurate medical knowledge about liver diseases and solve the demand for consultation.The main research contents of this paper are as follows:(1)The ALBERT-Bi LSTM-CRF model was used for medical entity recognition experiment.The crawling question and answer data set was used for training in ALBERT layer to obtain vector representation containing text semantic information.Then the semantic representation information of sentence vector is embedded into Bi LSTM layer for feature extraction,and the maximum probability label corresponding to each word is calculated.After that,the CRF layer decodes the module to obtain the optimal tag sequence corresponding to the input sequence,which is the result of named entity recognition.Finally,unlabeled data sets were used for model testing,and the results showed that the accuracy rate was 92.38%,the recall rate was 90.74%,and the F1 value was 91.55%.Besides,compared with other mainstream entity recognition models,the training time was shortened and the occupation of system resources was reduced.(2)To build a knowledge map of liver diseases.First,the semi-structured and unstructured data crawled from the medical website were cleaned and knowledge extracted,and stored in the file in json format.Then the entity,relationship and attribute types were defined and standardized,entity alignment and knowledge fusion were carried out,knowledge triplet was constructed and data was saved in csv files.Finally,Neo4 j database was used for storage.Construct the corresponding knowledge map.(3)Design and implement the liver disease question and answer system.The above research results are applied to the system construction.By solving the core tasks of semantic parsing and answer retrieval,the questions input by users are converted into Cypher query statements in the graph database,and then retrieved in the Neo4 j database and returned to the front-end interface.Finally,the system has realized the function of question and answer,map display,knowledge base management and user management.Through the above work,on the basis of the research on named entity recognition and knowledge map,an intelligent question and answer system for liver diseases is built.The system can answer users’ questions and display the knowledge map of related issues,and solve users’ demands for online consultation and medical information acquisition.