Research On The Construction And Application Of Causal Affair Graphs For Hot Topics

With the rapid development of the Internet and the popularization of mobile devices。,people have more and more access to hot topics,and every now and then,the topic with the highest degree of discussion occurs.Reports on various platforms increase the time for people to retrieve key information in a topic,and there is no intuitive expression of the cause and effect of an incident and protective measures.Topic events are real-time dynamic data,which will trigger different event results with time,place,person,and other factors.Studying causal logic between events is a difficult problem in natural text processing.In recent years,knowledge graph technology has been widely studied,but the focus is on static data.For the study of dynamic data,the concept of affair map was proposed,which is a kind of Logic Base of affair logic,which describes the evolution law between events.In view of the above problems,this article constructs a causal atlas for hot topics,uses event extraction technology to extract event elements,and uses event relationship extraction technology to obtain the logical relationship between events.Based on this,it implements topic query and intelligent question answering functions,and mainly completes work as follows:A framework for the construction of affair maps of hot topics is proposed.First,the data source is acquired using the Scrapy crawler framework,which is sorted in descending order of time.Based on hot topic titles,Kmeans unsupervised algorithm is used to divide the data source into several types of topics.The elements that define an event are composed of event participants,event trigger words,event occurrence location,event occurrence time,and event occurrence degree.Secondly,the sequence tagging method is used to transform the event extraction research into sequence tagging tasks,so as to extract event elements from hot topic events.The experiment compares three different sequence labeling models,and finds that the performance based on the BERT + Bi-LSTM + Attention + CRF model is the best,and the value of F1 on the test set reaches 91%.Then,the research on event relationship extraction,this paper respectively analyzes the explicit causality extraction based on dependency syntax analysis,the implicit relationship extraction based on event sentences and event pairs,and finally adopts an extraction model combining the characteristics of rules between events and Bi-GRU.The value of F1 on the test set reached 86%.Then based on the event pair consisting of event extraction elements,the semantically similarity calculation is used to obtain the two pairs of events with the highest score and 4876 causal event pairs extracted from the event relationship to construct <cause event,cause and effect,result event>,<event i,similarity,Event j> triplet,then the event is connected as an entity,the causal relationship between the event and the event,and the similarity relationship is stored in the Neo4 j graph database to realize the establishment of the logic of knowledge base of affairs and build the causal affairs of hot topics Atlas.Finally,based on the well-established hot topic causal affair atlas design,an affair atlas application system was developed,which realized the functions of topic query and intelligent question answering.In this paper,in the key technology of constructing the atlas of events,a variety of neural networks are combined in the event extraction technology to improve the accuracy of event element extraction.The concept of event pairs is proposed in the event relationship extraction,combining the characteristics of rules between events and the two-way short-term memory The causal events extracted by the model have higher accuracy.At the same time,Neo4 j graph database is adopted to overcome the shortcomings of insufficient depth query of traditional database.The causal atlas constructed based on this method saves the integration resources and construction time,improves the speed for users to obtain key information,and meets people’s needs for real-time hot topic understanding and protection measures.It is of great significance.