Research And Application Of Event Evolution Method Based On Event Logic Graph

While the rapid development of the Internet facilitates people’s lives,it also brings challenges to the management and control of public opinion.Negative public opinion events may have a negative impact on real life.Studying the analysis methods of event evolution has certain significance for maintaining social stability.Existing methods often have the problems of not paying high attention to the logical relationship between events and not strong interpretability.How to effectively reason and analyze the evolution of events is a problem that needs to be solved in the field of public opinion management and control.The Event Logic Graph is a new type of knowledge storage structure,which can record the evolution process of events that have occurred,reflect the law of event development,and has strong interpretability.In this paper,we propose an event evolution analysis method based on the Event Logic Graph.In view of the fact that there is no public Event Logic Graph data set,and the existing graph construction methods cover the problems of few logical relationships and weak event characteristics,a Multi-feature Fusion Event Logic Graph(MFFELG)construction method is proposed,which covers the five logical relationships of cause and effect,succession,transition,condition,concurrency,the coverage is more comprehensive.The entity information,event type,and semantic information of the event node are extracted as node attributes,which enhances the knowledge representation ability of the graph.Bi LSTM-CRF event extraction model fused with part-of-speech information is proposed,which improves the effect of event extraction through joint learning of text and corresponding parts of speech.After that,We propose an event evolution analysis method based on MFFELG,which including two processes of knowledge matching and knowledge reasoning.First,search for similar event nodes in the MFFELG according to the attribute characteristics and semantic information of the target event,and analyze the evolution direction of the target event according to the evolution law reflected by the graph.Secondly,in view of the problem that the graph may have incomplete links and still have unexplored event evolution relationships,a knowledge reasoning model Event Trans H is constructed to predict the development relationships of events that have not been fully recorded in the graph,and realize the evolution analysis of events.This paper conducts experimental analysis on event extraction model and knowledge reasoning model respectively,and demonstrates the event evolution analysis method proposed in this paper through examples.The results show that the Bi LSTM-CRF model fused with part-of-speech information achieves a good event extraction effect,the Event Trans H model improved the accuracy of affair knowledge reasoning,and the event attribute feature information had a positive impact on the model reasoning results.