Text Summarization using Fuzzy Relational Clustering Algorithm

Abstract

The amount of stored data is increasing day by day. The automatic document summarization is used to solve the problem of information overload. The general technique to generate the summary is to use clustering, where sentence clustering plays vital role. For clustering sentence level text fuzzy relational clustering algorithm is used. It considers graphical representation of data and graph centrality is used as likelihood; with the help of Expectation Maximization framework it determines cluster membership values and mixing coefficients. The results of algorithm when applied to sentence clustering task show that it is able to identify overlapping clusters of sentences which are related semantically. The output cluster membership values of the algorithm can be used to generate extractive summary of text. Experimental results on quotation dataset show that the algorithm performs far better as compare to the K-medoids algorithm. The summary produce by the fuzzy relational clustering algorithm is able to capture the main idea of the original text and it provides useful information to users.