Construction And Completion Of Knowledge Graph Of COVID-19 Cases

Since the spread of the new coronavirus in 2019,my country has taken many effective measures to prevent and control the epidemic.However,the epidemic is not over yet,so the research on strengthening the prevention and control of the new crown epidemic is still in progress.Due to the variety of transmission routes and strong camouflage of the new coronavirus,epidemic prevention and control are mainly carried out by tracking the spatiotemporal data of cases and identifying close contacts.Recently,knowledge graphs have been applied in many fields,including question answering systems,search engines,intelligent speech,and recommendation systems.With the development of knowledge graphs,this paper considers applying knowledge graphs in the field of epidemic prevention and control.According to the case trajectory information,a new crown epidemic knowledge graph composed of new crown epidemic cases and POI data of cities is constructed,and the missing infection events are supplemented.It can realize information traceability,infer high-risk locations,discover close contacts,etc.The new crown epidemic knowledge graph constructed in this paper is a time knowledge graph.Unlike most of the existing knowledge graphs,the time knowledge graph is a knowledge graph with time information,and the knowledge in the time knowledge graph is usually temporarily valid.The reason why the new crown epidemic knowledge graph is constructed as a time knowledge graph is mainly because in the new crown epidemic,a lot of information is related to time,such as the time of contact with cases and the time of visiting a high-risk area.Determine the probability of contracting the new coronavirus.This paper also visualizes the constructed new crown epidemic knowledge graph,in order to make the information more intuitive.When building the knowledge graph of the new crown epidemic,we found that the case trajectory information publicly released on the Internet was incomplete,which was caused by many uncontrollable factors.Incomplete data will lead to the lack of knowledge in the knowledge graph,which will bring difficulties to the subsequent use of the new crown epidemic knowledge graph for new crown prevention and control research.It is necessary to complete the constructed knowledge graph of COVID-19 so as to find out the parts of the knowledge graph that need to be supplemented.There are many existing knowledge graph completion models,such as knowledge representationbased models,path reasoning-based models,and reinforcement learning-based models,etc.,but these models are all used for static knowledge graph completion work,static knowledge graphs.That is,a knowledge graph without time information.The COVID-19 knowledge graph constructed in this paper is a temporal knowledge graph.Although static knowledge graphs are widely used in relational reasoning and downstream tasks,they cannot realistically model knowledge and facts that are only temporarily valid.Also,most existing temporal knowledge graph completion models extend static knowledge graph embeddings,they do not take full advantage of the temporal knowledge graph structure,and thus lack consideration of temporally relevant events that already exist in the local neighborhood of the query,and in the inference Lack of interpretability in the process.Therefore,this paper proposes a new model for the completion of the knowledge graph of the new crown epidemic: T-WGPR.The model is mainly divided into two parts,one is the encoder and the other is the decoder.In the encoder of T-WGPR,the query-specific substructure of the COVID-19 knowledge graph is encoded by using a graph convolutional neural network with attention mechanism to focus on the temporal displacement between each event and the query timestamp,and then through the graph Up-propagating attention to the decoder to perform path-based inference.Finally,this paper uses the constructed knowledge map of the new crown epidemic to conduct experiments on the proposed knowledge map completion model.The experimental results show that the proposed model is compared with the existing knowledge map completion models in each evaluation index.The average increase is about 10%.