Frame Semantics for Stance Classification

Proceedings of the Seventeenth Conference on Computational Natural Language Learning, pages 124–132, Sofia, Bulgaria, August 8-9 2013. c⃝2013 Association for Computational Linguistics Frame Semantics for Stance Classification Kazi Saidul Hasan and Vincent Ng Human Language Technology Research Institute University of Texas at Dallas Richardson, TX 75083-0688 {saidul,vince}@hlt.utdallas.edu Abstract Determining the stance expressed by an author from a post written for a two-sided debate in an online debate forum is a relatively new problem in opinion min- ing. We extend a state-of-the-art learning- based approach to debate stance classifica- tion by (1) inducing lexico-syntactic pat- terns based on syntactic dependencies and semantic frames that aim to capture the meaning of a sentence and provide a gen- eralized representation of it; and (2) im- proving the classification of a test post via a novel way of exploiting the information in other test posts with the same stance. Empirical results on four datasets demon- strate the effectiveness of our extensions. 1 Introduction Given a post written for a two-sided topic in an online debate forum (e.g., “Should abortion be al- lowed?”), the task of debate stance classification involves determining which of the two sides (i.e., for or against) its author is taking. For example, a stance classification system should determine that the author of the following post is anti-abortion. Post 1: Abortion has been legal for decades and no one seems to have a problem with it. That’s ridicu- lous! There are millions of people in the world who would love to have children but can’t. Previous approaches to debate stance classifica- tion have focused on three debate settings, namely congressional floor debates (Thomas et al., 2006; Bansal et al., 2008; Balahur et al., 2009; Yesse- nalina et al., 2010; Burfoot et al., 2011), company- internal discussions (Murakami and Raymond, 2010), and online social, political, and ideologi- cal debates in public forums (Agrawal et al., 2003; Somasundaran and Wiebe, 2010; Wang and Ros´e, 2010; Biran and Rambow, 2011; Hasan and Ng, 2012). As Walker et al. (2012) point out, debates in public forums differ from congressional debates and company-internal discussions in terms of lan- guage use. Specifically, online debaters use color- ful and emotional language to express their points, which may involve sarcasm, insults, and question- ing another debater’s assumptions and evidence. These properties can potentially make stance clas- sification of online debates more challenging than that of the other two types of debates. Our goal in this paper is to improve the state of the art in stance classification of online de- bates, focusing in particular on ideological de- bates. Specifically, we present two extensions, one linguistic and the other extra-linguistic, to the state-of-the-art supervised learning approach to this task proposed by Anand et al. (2011). In our linguistic extension, we induce patterns from each sentence in the training set using syntactic depen- dencies and semantic frames that aim to capture the meaning of a sentence and provide a general- ized representation of it. Note that while Anand et al.’s lexico-syntactic approach …

Content truncated to 500 words