Extracting Condition-Opinion Relations Toward Fine-grained Opinion Mining

Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 622–631, Lisbon, Portugal, 17-21 September 2015. c⃝2015 Association for Computational Linguistics. Extracting Condition-Opinion Relations Toward Fine-grained Opinion Mining Yuki Nakayama Atsushi Fujii Department of Computer Science Graduate School of Information Science and Engineering Tokyo Institute of Technology {nakayama.y.aj@m,fujii@cs}.titech.ac.jp Abstract A fundamental issue in opinion mining is to search a corpus for opinion units, each of which typically comprises the evalua- tion by an author for a target object from an aspect, such as “This hotel is in a good location”. However, few attempts have been made to address cases where the validity of an evaluation is restricted on a condition in the source text, such as “for traveling with small kids”. In this paper, we propose a method to ex- tract condition-opinion relations from on- line reviews, which enables fine-grained analysis for the utility of target objects de- pending the user attribute, purpose, and situation. Our method uses supervised machine learning to identify sequences of words or phrases that comprise conditions for opinions. We propose several features associated with lexical and syntactic infor- mation, and show their effectiveness ex- perimentally. 1 Introduction Reflecting the rapid growth in the use of opin- ionated texts on the Web, such as comments on news articles and customer reviews, opinion min- ing has been explored to facilitate utilizing opin- ions mainly for improving products and decision- making purposes. While in a broad sense opin- ion mining refers to a process to discover useful knowledge latent in a corpus of opinionated texts, fundamental issues involve modeling an unit of opinions and searching the corpus for those units, each of which typically comprises the evaluation by an author for a target object from an aspect. Other elements, such as when the opinion was sub- mitted, can optionally be included in an opinion unit. We take the following review sentence as an example opinionated description. (1) I think hotel A offers a reasonable price if you take a family trip with small kids. From the above example, existing methods (Pang and Lee, 2008; Seki et al., 2009; Jin et al., 2009; Zhao et al., 2010; He et al., 2011; Liu and Zhang, 2012; Liu et al., 2013; Yang and Cardie, 2013; Liu et al., 2014) are intended to extract the following quintuple as an opinion unit. Target = “hotel A”, Aspect = “price”, Evaluation (Polarity) = “reasonable” (positive), Holder = “I (author)”, Time = N/A Depending on the application, “Evaluation” can be any of a literal opinion word (e.g., “reasonable”), a polarity (positive/negative), or a value for multi- point scale rating. Given those standardized units extracted from a corpus, it is feasible to overview the distribu- tion of values for each element or a combination of elements. For example, those who intend to im- prove the quality of hotel A may investigate repre- sentative values for “Aspect” in the units satisfy- ing “Target=hotel A & Polarity=negative”, while those who look for accommodation may collect …

Content truncated to 500 words