The Method And Its Application Of Speech Text Analysis Based On Multi Document Summarization

Group discussion support system is an important part of the hall for workshop of meta synthesis,which serves to support group members to exchange their views online and to analyze them comprehensively.In the course of discussion,a large amount of speech texts will be generated.If these texts are not analyzed and processed property,it is not conducive for participants to understand the logical relationship among different speech content.At present,the common method of speech text analysis is to use argumentation theory in artificial intelligence to determine the acceptability or consensus value of statements by analyzing the relationship among arguments or the strength of attack or support.However,in the process of text-based discussion,there are no clear attack or support relationships among speech texts,and the group consensus is hidden in participants' speeches,so it is necessory to analyz the content of speech texts.In terms of this issue,this paper puts forward a method of analyzing the content of multi-discussion text and extracting the consensus information from speech texts,and applies the method to the group discussion support system.The specific work as follows:1.A method of speech text analysis based on multi-document summarization is proposed.Firstly,the text preprocessing is carried out;and then the heuristic clustering algorithm is used to cluster and analyze the speech texts.Secondly,the multi-document summary algorithm is used to extract the summary of each text cluster.Finally,the D3 technology is used to show the results of text clustering and multi-document summary.2.On the basis of analyzing the classical Text Rank algorithm,an improved Text Rank text digest algorithm is proposed.The core idea of this algorithm is to add the information such as sentence position,keyword sentence and sentence length to the sentence weight calculation.At the same time,the similarity calculation method of the improved sentence is given and the experiment proves that the algorithm is feasible.3.A web-based discussion text analysis subsystem is developed.The subsystem is integrated into an existing group discussion support system.Experimental analysis is carried out with a set of speech texts from a discussion case,and different clustering results and multi-document summaries are obtained by adjusting the threshold of clustering algorithm.Experimental results show that the proposed speech text analysis method based on multi-document summarization is feasible,which can effectively promote the convergence of group thinking,and improve the efficiency of discussion.