Temporal Relationship Recognition And Its Application For Emergency Events

In the era of information explosion,a large number of news events will be produced every day.As the basic unit of human knowledge,the relevant research on events has gradually attracted people's attention.Events often do not exist in isolation,they are related to each other through a certain chronological order.By identifying the temporal relationship of the events in the news text,the semantic relationship of the events in the text in time can be obtained,which helps people to understand the content of the text faster and better.At the same time,the use of the temporal relationship between events is of great significance in the research work of automatic question answering system,automatic abstract extraction,etc.This paper focuses on the identification of Chinese emergencies in time sequence.Study lacks is used to identify Chinese temporal relationships between the features,the traditional weak effect on events that are farther apart,identification results will appear in some problems such as temporal logic conflict,this paper proposes a event timing relationships based on the characteristics and rules of constraint recognition methods,and on the basis of the event-based automatic line is used to study of temporal relations.The research mainly includes the following two aspects:(1)Event timing relationship recognition based on multiple features and rule constraints.In order to identify the temporal relationship of emergencies,we firstly preprocessed the text of the corpus using the language processing tool LTP of Harbin Institute of Technology,and then performed the steps of word segmentation,part of speech tagging,dependency parsing and so on after the text of XML annotation format was retrieved.On this basis,the basic features of sentences and words in the text were obtained.Then,according to the characteristics of Chinese language expression,we further extract the event elements,special words,causal markers and trigger words,and use the maximum entropy model to identify the temporal relationship between events.Finally,the idea of integer linear programming was proposed to solve the problem that there would be some temporal logic conflicts in the identification results,and the custom rules,such as connectors,event types and time information between event pairs,were taken as constraints,and the identification results were further optimized by combining with the improved objective function.The experimental results show that the recognition method combining multiple features and rule constraints can improve the recognition effect of the time series relationship between events.(2)Automatic abstract extraction based on event timing relational network.Combined with the idea of graph model,it is proposed to construct the event sequence relation network by using the time sequence relation between events to extract the s.Firstly,the annotated event triggering words and event relations in the corpus are used to construct the text representation model of event timing relations with the event triggering words as the node and the event timing relations as the edge.Then the PageRank algorithm is used to calculate the weight of each node in the network corresponding to the event,and the final weight adjustment is made by the similarity between the event sentence and the text title and the content of effective words.Edge finally by using the improved MMR(maximum)algorithm for redundant processing in order to increase the diversity,select the MMR's highest sentence added to the collection,selection until the target number of the words,and according to the events in the development of temporal relations,the event of a set of words to the final sorting,summary of the text.Relevant experiments show that the automatic abstract extraction method based on event timing relation network has achieved good experimental results in Rouge evaluation.Figure[9]Table[15]Reference[60].