An Effective Sentence Ordering Approach For Multi-Document Summarization Using Text Entailment

Abstract

With the rapid development of modern technology electronically available textual information has increased to a considerable amount. Summarization of textual inform ation manually from unstructured text sources creates overhead to the user, therefore a systematic approach is required. Summarization is an approach that focuses on providing the user with a condensed version of the origina l text but in real time applicat ions extended document summarization is required for summarizing the text from multiple documents. The main focus of multi - document summarization is sentence ordering and ranking that arranges the collected sentences from multiple document in order to gene rate a well - organized summary. The improper order of extracted sentences significantly degrades readability and understandability of the summary. The existing system does multi document summarization by combining several preference measures such as chronology, probabilistic, precedence, succession, topical closeness experts to calculate the preference value between sentences. These approach to sent ence ordering and ranking does not address context based similarity measure between sentences which is very ess ential for effective summarization. The proposed system addresses this issues through textual entailment expert system. This approach builds an entailment model which incorpo rates the cause and effect between sentences in the documents using the symmetric measure such as cosine similarity and non - symmetric measures such as unigram match, bigram match, longest common sub - sequence, skip gram match, stemming. The proposed system is efficient in providing user with a contextual summary which significantly impro ves the readability and understandability of the final coherent summa.