Research On Automatic Summarization Method Based On Event Elements Network For Multiple Topics Text

With the continuous development of automatic summarization technique in the field of natural language processing,more and more researchers begin to focus on how to obtain the summarization of a text accurately and efficiently.In recent years,scholars pay high attention to the concept event in the field of natural language processing.But there no more research work on automatic summarization based on event in the aspect of automatic summarization technique.Therefore,this paper studies the extraction method of automatic summarization based on the event elements network from the angle of event elements on the size of event.There are two parts included in this paper.1.Research on default completion of event elements.Event can be defined as a 6-tuple formally.One or several event elements may be omitted,but usually appears in other events.Our research work on event elements default completion is mainly about object elements,time elements and environment elements(place elements)default completion.On the basis of observation and analysis on each event element,this paper uses the method of combining the similar event trigger table and the rules based on the context structure to implement the default time elements and place elements,the default object elements,but the way of the used above default completion method is different.Our experiments have achieved better results about the default completion of the event elements above.2.Research on extraction method of automatic summarization based on event elements network.After completing the default event elements,we firstly construct an event elements network by using event elements.Then utilizing relative knowledge of graph theory,we calculate each node importance of the event elements network in order to obtain each event element importance.So each sentence importance is obtained indirectly.Then we use the community partition method to solve the problem of multiple topics partition.Lastly,we deal with the redundant information according to the event similarity,and then output the summary sentences to obtain summarization of the text according to the appearance order of them in the original text.We select the CEC as the experimental corpus,and the experimental results show that both the precision and recall of the proposed automatic summarization method in this paper have better performance.