Research On The Identification Method Of Chinese News Event Association Relationship Based On The Event Chain Structure

Information extraction technology as a hot research topic in the field of natural language processing has gradually become the research focus of domestic and foreign researchers in the last ten years.As an important research branch of information extraction,event correlation recognition is of great significance.However,at present,researches on event relationship recognition both at home and abroad focus on the identification of specific relationships,such as causality,and there are relatively few studies on the identification of event correlation.Event relevance recognition by analyzing the semantic and structural characteristics of the text gives the relevant or unrelated decision to the text fragments that describe different events.This paper is based on the structure of news event chain to study the relationship between events.An event chain is a data structure that holds information about the related attributes of the events.Studying the relationship of news events on the basis of the event chain structure can not only accurately locate the relevant event pairs,narrow the range of events with associated relationship,and eliminate redundant and useless judgment,but also greatly improve the accuracy of judgment.Based on event chain structure,this paper focuses on corpus construction,event related character recognition and Event Association recognition.The main Achievements are as follows:(1)Building an event associated character corpus.This paper takes ACE2005 Chinese corpus and predicate argument corpus as the prototype.7 kinds of event correlation characteristics are manually labeled.The annotation results were analyzed in detail,and stored in the database to construct corpus of event correlation characters.(2)Event related character recognition.Constructing the maximum entropy classifier by selecting the feature space,the event association features are identified.On the basis of this,a method to identify the correlation characteristics of events.(3)According to the results of the previous research,two algorithms for calculating the correlation between news events and identifying the correlation of news events are proposed.(4)Integrating the above 3 research results,a prototype system of news event correlation recognition based on event chain structure has been developed.