Research On Event Element Recognition For Emergencies

In recent years,all over the world frequent occurrence of all kinds of emergencies,such as: earthquakes,fires,terrorist attacks,the emergence of a large number of emergencies so that serious redundancy of information on the Internet,it is difficult for people to obtain useful information,event extraction is driven by this phenomenon quietly rise.Event extraction is an important research direction in the field of information extraction,which mainly includes two aspects: event identification and event recognition.Event recognition mainly identifies the event trigger words,and the identification of event elements includes time factor,location element and object element.This paper focuses on the identification of event elements.(1)In natural language,the time element of an event reflects the time when the event occurred.Time can be divided into specific time and relative time.The place element reflects the specific location where the event occurred,and there are no events existing apart from the place element.In this paper,the improved K-means algorithm and Prefix Span algorithm are used to identify the event time and location features.K-means algorithm is an unsupervised clustering algorithm,but K-means algorithm has the problem that the number of clusters can not be determined.Therefore,Canopy algorithm is introduced to finish the "rough clustering" of source data first and apply the results of "rough clustering" to K-means algorithm.After the clustering is completed,the clustering results are deeply mined by prefix span algorithm,and the results of mining are artificially processed to obtain the recognition rules of time and place elements,and then to recognize the event time and location elements.The experiment obtains the more ideal effect.(2)Identification of event object elements.The object element refers to the main participant of the event,which can be divided into two types: the Subject object(the action initiator)and the object target(the action recipient),this paper proposes a method based on the dependency parsing to identify the event object elements.This method first studies the linguistic representation law of the object elements in the event,and defines the heuristic rules used to judge the existence of the object elements in the event,then on this basis,the event of the object elements is identified by the method of dependency parsing,and finally the effect is validated by experiment.