A New Method of Using Contextual Information to Infer the Semantic Orientations of Context Dependent Opinions

Abstract

many researches focus on semantic analysis in opinion mining and get effective results. However, dealing with context dependent opinions is still a challenge. Existing methods have used linguistic rules to cope with this problem, however, when the opinion is irrelevant to its adjacent sentences, the linguistic rules will not work. These special opinions, in this paper, called context indistinct-dependent opinion. We propose a method to predict the orientation of these special opinions. The main idea of our approach is when the review an opinion lies in cannot provide enough contextual information to determine the orientation of opinion, we resort to other reviews discussing the same topic to mine useful contextual information, then use semantic similarity measures to judge the orientation of opinion. Our approach is innovative and experiment shows that the proposed technique is highly effective.