Use of fuzzy logic and wordnet for improving performance of extractive automatic text summarization

Abstract

Text Summarization produces a shorter version of large text documents by selecting most relevant information. Text summarization systems are of two types: extractive and abstractive. This paper focuses on extractive text summarization. In extractive text summarization, important sentences are selected based on certain important features. The importance of some extractive features is more than the some other features, so they should have the balance weight in computations. The purpose of this paper is to use fuzzy logic and wordnet synonyms to handle the issue of ambiguity and imprecise values with the traditional two value or multi-value logic and to consider the semantics of the text. Three different methods: fuzzy logic based method, bushy path method, and wordnet synonyms method are used to generate 3 summaries. Final summary is generated by selecting common sentences from all the 3 summaries and from rest of the sentences in union of all summaries, selection is done based on sentence location. The proposed methodology is compared with three individual methods i.e. fuzzy logic based summarizer, bushy path summarizer, and wordnet synonyms summarizer by evaluating the performance of each on 95 documents from standard DUC 2002 dataset using ROUGE evaluation metrics. The analysis shows that the proposed method gives better average precision, recall, and f-measure.