SUMMARIZATION OF TEXT USING FUZZY RELATIONAL CLUSTERING

Abstract

Information overload is a major problem in the modern digital world. It is difficult to retrieve the relevant content from the billions of documents. Moreover, the mobile devices have restricted memory, display screen and processing power. The mobile users prefer to analyse the summarized report, if it is relevant to their requirement, the user may observe it deeper. However, it is difficult to manually summarize the large documents of text. Sentence level clustering can be employed to perform automatic summarization task. Most of the sentence similarity measures do not represent sentences in common metric space. Hence, the conventional fuzzy clustering approaches based on prototypes are not appropriate for sentence level clustering. The Expectation Maximization (EM) framework is used to perform lexrank analysis on relational input data iteratively. The proposed work computes the centroid sentence of each cluster to automatically generate the summary which in turn reduces the manual processing. The integration of automatic summarization process to mobile devices is deployed by using Extensible Markup Language (XML). Since XML is originally designed to support large scale electronic publishing of documents, it is a flexible interface for mobile devices inorder to relieve the reading burden of the users. Experimental evaluation on the famous quotations dataset shows that the fuzzy relational clustering algorithm is capable of extracting the semantically related important sentences for generating the summarized content of original source.