Research On The Construction Of An Event Evolution Graph For The COVID-19

Knowledge graph focus on various entities or concepts existing in the objective world and their relationships,while ignoring the knowledge of evolutionary laws between events.The nodes in the event evolution graph are highly generalized events,and the edges are the evolutionary relationships between events,such as causal relationships,succession relationships,upper-lower relationships,and time-series relationships.The development of natural language processing technology has improved the accuracy of information extraction,and it has also made it possible to automatically mine event knowledge from text and build an event evolution graph.This paper starts with the "COVID-19" event in the field of public health emergencies,and conducts research on the key technologies for the construction of event evolution graph.The specific contents include COVID-19 event extraction,event relation extraction,event generalization,and graph visualization to form an event evolution graph.The acquisition of affinity knowledge is the basis for constructing event evolution graph.The current English corpus mainly includes the ACE Conference Open Data Set,the Chinese Emergencies CEC Corpus,etc.These corpora are relatively scattered in distribution.The form and specification of the annotation needs to be further converted from the word level to the sentence level and then to the chapter level.This article selects official media articles on the 2019 new COVID-19 from People’s Daily Online,China News Network,Lilac Garden and Guangming.com,etc,totaling 5,000 articles,using Python to write crawlers,crawl the real-time broadcast of the epidemic news on the platform,and use Harbin Institute of Technology LTP tool for word segmentation,part-of-speech tagging and sentence parsing to obtain a semantic role dictionary,and use the joint method based on trigger wordevent type comparison table to extract event trigger words and type identification;use event element extraction based on dependency syntax analysis and matching rules and maximum entropy classifier method.In this paper,event causality extraction is divided into explicit and implicit relation extraction,pattern matching method is used for explicit causality,and deep learning extraction method is used for implicit causality.The experimental results on the corpus demonstrate the effectiveness of the method.Events are the core elements of the event evolution graph.In order to better describe the event evolution graph network,it is necessary to generalize specific events into more abstract events.This paper integrates the generalization method based on event similarity calculation,so that the events obtained after generalization are integrated into the graph network can better expand the density of the graph network,so that the event evolution graph can be better applied to other tasks.Based on the results of the above research plans,this paper designs and implements an event evolution graph for the COVID-19 event,and verifies the feasibility of the COVID-19 event evolution graph construction method based on the corpus.