Research On COVID-19 Knowledge Graph Construction And Intelligent Question Answering Method

Knowledge graph is a structured semantic knowledge base.It expresses the relationship between knowledge intuitively and clearly through graph database,which can make knowledge have better sharing and reusability.Constructing knowledge graph can meet the needs of large-scale data mining and management.Knowledge Graph Question Answering System(KGQA)selects several entities or attribute values from a given knowledge graph as the answer to the question.Compared with traditional search engine technology,KGQA does not require users to manually screen documents,and relies on high-quality knowledge graphs for data support,so it has better stability and accuracy.Corona Virus Disease 2019(COVID-19)has gradually spread to all over the world since its discovery,posing a serious threat to global public health security.Aiming at people ’s information needs for the novel coronavirus,and in order to overcome the problems of information fragmentation and low knowledge utilization of the novel coronavirus,this paper first constructs the COVID-19 knowledge graph and proposes a knowledge graph question answering method.On this basis,a new coronavirus knowledge graph question answering system was developed to help users better grasp the relevant knowledge in the field of new coronavirus.Based on the above analysis,the main research contents of this paper are as follows :(1)Automatic construction of COVID-19 knowledge graph.Based on the bottom-up knowledge graph construction technology,the Scrapy crawler technology is used to crawl the knowledge related to the novel coronavirus from the medical website,and then the IDCNN model is used for relationship extraction.Then we use a method to automatically build a knowledge graph in a relational database.This method identifies the database type by the database table identification method.According to the database type,the relational schema conversion method can automatically generate the ontology of the Owl file format.The RDF file is generated by the R2 RML data mapping technology,and the knowledge graph is successfully generated after the RDF file is imported into Neo4j;finally,it is stored in the Neo4 j graph database to realize the hierarchical interconnection and semantic processing ability in the form of graph structure,and intuitively display the systematicness and relevance of knowledge.(2)Based on the constructed COVID-19 knowledge graph,a knowledge graph question answering method is proposed to realize the retrieval and utilization of COVID-19 knowledge graph.The knowledge graph question answering method proposed in this paper first uses the ALBERT-Bi LSTM-CRF model to perform named entity recognition on natural language questions,and then uses an entity linking algorithm combining naive Bayesian classifier and edit distance calculation to generate question candidate triples in the knowledge graph as the answer result.The ALBERT-Bi LSTM-CRF named entity recognition model is compared with other mainstream models.The experimental results show that the accuracy,recall rate and F1 of named entity recognition are greatly improved.(3)Design and implement a COVID-19 knowledge graph intelligent question answering system.The system has undergone detailed requirements analysis and architecture design.It uses a front-end and back-end separation design.The front-end is built using Vue + Bootstrap,and the back-end uses Python to encapsulate interfaces such as database query,question answering system retrieval,and knowledge base construction.The system includes three modules:user management,knowledge graph question answering and background administrator management.It can realize the basic functions of user login,registration,personal information management,visual question answering,system feedback and so on.Aiming at the problem of slow user input,voice input function is added to the system to help users manage and use information more conveniently.It can meet the practical application requirements.