Utilizing local sentiment information in sentiment topic detection for online news documents

Abstract

Facts and opinions are two main types of textual information on the Web. Though current research focuses on the analysis of factual information in text, understanding sentiment and opinions expressed in text is getting important to understand text. Sentiment topic detection summarizes sentiment expression of the same target distributed over documents. Researches on sentiment topic detection have found sentiment topic by analyzing topic over documents and associating sentiment expression to topics. This paper proposes an approach of sentiment topic detection that utilizing local sentiment expression to detect sentiment topic. Sentiment expressions in each document are analyzed and documents are grouped by sentiment expressions of them. Through MPQA dataset, we found proposed approach reduces clustering error of sentiment topics compared.