Research On Scene Portrait Of Railway Emergency Based On Event Evolution Graph

The portrait of railway emergency can provide data for the analysis of the cause of railway accident,provide basis for the decision of accident rescue,and provide reference for the determination of accident liability,which is of great significance to the railway emergency rescue guidance.At present,the portrait of railway emergencies at home and abroad mainly uses keyword description and on-site three-dimensional digital portrait method,which can’t combine the entity and attribute of railway event,and can’t describe the relationship of the event.In view of the above problems,this paper adopts the method based on the event evolutionary graph,extracts the event representation and labeling,event and event relationship extraction and generates the event evolutionary graph three part to make a portrait of the railway emergency,focuses on the ABC event representation model(REABCO)of railway emergency,railway emergency annotation based on BIO joint annotation,extraction of explicit event relations based on syntactic dependency,and implicit events based on bidirectional LSTM model(BERT+Bi-LSTM+Attention+Softmax)with Attention mechanism.Finally,the software Neo4 j was used to generate the theory graph.The specific content is as follows:In this paper,the portrait of railway emergency based on the event evolutionary graph is divided into three parts: event representation and event labeling,event extraction and event causality extraction,and finally,the generation of the event evolutionary graph.In terms of event representation and event marking,this paper defines the ABC event representation model for railway emergency according to the analysis of event division in the field of railway emergency case and emergency,and further perfects the ABC event representation model according to the characteristics of railway emergency,enriches the situation layer,and constructs the event representation model REABCO for railway emergency.According to REABCO,the railway emergency marking method based on BIO joint labeling was selected,and the railway emergency information was labeled with data.The labeling method takes the excavated railway emergency text information as the original information on the social media platform headed by weibo,identifies both the entity and the trigger word from the unstructured text of the social platform according to the BIO labeling method and extracts the sentence containing the description of the railway emergency,which enables the computer to identify the event elements,so as to realize the data labeling of the railway emergency.In terms of event extraction and event causality extraction,this article divides event extraction into two parts.One type converts part of the event extraction task into an annotation task during the event annotation process,and the other is based on Bidirectional long-short-term memory model incorporating memory mechanisms(BERT+Bi-LSTM+Attention+Softmax).The extraction model for event extraction.Event sentences containing time causal trigger words are more suitable for the first type of method to extract events,while event sentences that do not contain causal trigger words need another method to achieve event extraction.For the extraction of causality in railway emergencies,according to the characteristics of explicit causality and implicit causality,extraction methods based on pattern matching and neural network are adopted respectively.Distinguish the explicitness and implicitness of causality by analyzing whether the event sentence contains trigger words of causality.For event sentences that contain causal triggers,the causal relationship is extracted according to the syntactic structure;for event sentences that do not contain causal triggers,the events in the sentence are analyzed in groups.Finally,generate event pairs.Neo4j is used in this paper to construct the event evolutionary graph of railway emergencies.Neo4 j is a high-performance graph database,which can build a railway emergency database through Neo4 j,and can build a visual graph for the convenience of experts to inquire.Based on social media platform and information as data sources,this paper generates the "3·30" Chenzhou bullet train derailment accident in Hunan and the "7·23" Yongwen Line railway accident.The experimental results show that the event evolutionary graph can reflect the key information and relationship of the railway emergency and describe the development process of the incident,so as to realize the accurate portrait of the railway emergency.The research results show that event evolutionary graph method can better complete the portrait work of railway emergencies,show the main elements of railway emergencies,and describe the relationship between events and the development process of railway emergencies.Compared with other representation models,REABCO subdivides railway emergencies,which can better represent railway emergencies and refine the nodes of the graph.At the same time,the method of syntactic dependency and neural network can improve the accuracy of event extraction,so as to improve the accuracy of the graph.