Recently, Studies on opinion mining have shown great value in practical applications. Existing works mainly focus on the components of opinion, such as lexicon orientation, opinion holder and opinion targets. This works can help people understand the overall evaluation on a specific topic or object from certain source. However, the reasons and/or consequences behind an opinion can be varied. Ignorance of this valuable information may lead users’ confusion and misunderstanding. According to our review, the explanation of opinion is not taken into consideration in the current literature. To address this challenge, in this paper, we focus on the explanation of reason and/or consequence behind an opinion and propose a method to extract the explanation of opinion automatically based on characteristics of Chinese online review. The accomplishments of this work include the following aspects: a) We review the the main tasks and approaches of existing work on opinion mining and introduce the concept of opinion explanation. Through analyzing their limitations of output, we clarify the necessity of extracting the explanation of opinion. b) We design rules to extract opinion explanation with a causal indicator in Chinese online reviews. We adopt word pairs as effective features and extract opinion explanation without a causal indicator. Also we optimize the extracting results of cases with causal indicators by ranking rules with word pairs. c) We adopt thesaurus to calculate the semantic similarity of words, and improve the extraction performance when taking word pairs as features. d) We conduct experiments on a Chinese hotel review corpus to verify the effectiveness and the feasibility of our proposed method. Our algorithm for extracting explicit and implicit cases reaches a fairly good accuracy. This thesis research can help improving user experiences for decision making and expanded the output of existing opinion researches.
修改评论