Text Summarization Using Sentiment Analysis for DUC Data

Abstract

Text Summarization is a way of determining the key concepts that are covered in the given text under consideration. Various techniques have been presented in literature and many are used in commercially available systems. The primary aim of the present work is to find efficacy of using sentiments for text summarization. In this work we present a computationally efficient technique based on sentiments of key words in text for the purpose of summarization. Sentiment analysis is already being used in various domains for analysis of large scale text data interpretation and opinion mining. The present work shows that sentiment analysis can also be used efficiently for the purpose of text summarization. We have tested our results on the standard DUC2002 datasets, and compared our results with different summarization approaches, viz. Random indexing based, LSA based, Graph based and Weighted graph based methods for different percentages of summarization. The proposed scheme is found to be efficient, in particular for 50% summarization.