Phrase Dependency Parsing for Opinion Mining

Proceedings of the 2009 Conference on Empirical Methods in Natural Language Processing, pages 1533–1541, Singapore, 6-7 August 2009. c⃝2009 ACL and AFNLP Phrase Dependency Parsing for Opinion Mining Yuanbin Wu, Qi Zhang, Xuanjing Huang, Lide Wu Fudan University School of Computer Science {ybwu,qi zhang,xjhuang,ldwu}@fudan.edu.cn Abstract In this paper, we present a novel approach for mining opinions from product reviews, where it converts opinion mining task to identify product features, expressions of opinions and relations between them. By taking advantage of the observation that a lot of product features are phrases, a con- cept of phrase dependency parsing is in- troduced, which extends traditional depen- dency parsing to phrase level. This con- cept is then implemented for extracting re- lations between product features and ex- pressions of opinions. Experimental eval- uations show that the mining task can ben- efit from phrase dependency parsing. 1 Introduction As millions of users contribute rich information to the Internet everyday, an enormous number of product reviews are freely written in blog pages, Web forums and other consumer-generated medi- ums (CGMs). This vast richness of content be- comes increasingly important information source for collecting and tracking customer opinions. Re- trieving this information and analyzing this con- tent are impossible tasks if they were to be manu- ally done. However, advances in machine learning and natural language processing present us with a unique opportunity to automate the decoding of consumers’ opinions from online reviews. Previous works on mining opinions can be di- vided into two directions: sentiment classification and sentiment related information extraction. The former is a task of identifying positive and neg- ative sentiments from a text which can be a pas- sage, a sentence, a phrase and even a word (So- masundaran et al., 2008; Pang et al., 2002; Dave et al., 2003; Kim and Hovy, 2004; Takamura et al., 2005). The latter focuses on extracting the el- ements composing a sentiment text. The elements include source of opinions who expresses an opin- ion (Choi et al., 2005); target of opinions which is a receptor of an opinion (Popescu and Etzioni, 2005); opinion expression which delivers an opin- ion (Wilson et al., 2005b). Some researchers refer this information extraction task as opinion extrac- tion or opinion mining. Comparing with the for- mer one, opinion mining usually produces richer information. In this paper, we define an opinion unit as a triple consisting of a product feature, an expres- sion of opinion, and an emotional attitude(positive or negative). We use this definition as the basis for our opinion mining task. Since a product review may refer more than one product feature and ex- press different opinions on each of them, the rela- tion extraction is an important subtask of opinion mining. Consider the following sentences: 1. I highly [recommend](1) the Canon SD500(1) to anybody looking for a compact camera that can take [good](2) pictures(2). 2. This camera takes [amazing](3) image qualities(3) and its size(4) [cannot be beat](4). The phrases underlined are the product features, marked with square …

Content truncated to 500 words