Semantic based Sentence Ordering Approach for Multi-Document Summarization

Abstract

With the rapid growth of online information which is unstructured in nature poses a great challenge to the text mining algorithms to retrieve useful and meaningful information in an efficient way. However larger amount of data are readily available, it is very difficult to access the required information at the right time and also in the most appropriate form. Therefore a systematic approach called multi-document summarization is required to generate a summary about particular topic. The main focus of document summarization is sentence ordering and ranking. The existing system for sentence ordering deals with the measures such as chronology, topical, precedence and succedence experts. The main drawback of existing system is, it does not address the semantic relationship between the sentences in the summary which is necessary to create a meaningful summary. The proposed system addresses the semantic relationship between sentences in the summary using wordnet synsets. This system builds an entailment model which infer the logical relationship among the sentences when arranging the sentences in the summary. Graph method is used for ranking the sentences, where nodes represents the sentences and the edges represents the preference value of one sentence over another sentence. The proposed system provides an efficient summary which is considerably increases the meaningfulness of the final summary and also typically recovering the user from the information overload problem by giving quick and efficient access to required information. Index Terms-Multi-document summarization, sentence ordering, sentence ranking, semantic expert, text entailment expert.