A Step-wise Usage-based Method for Inducing Polysemy-aware Verb Classes

Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics, pages 1030–1040, Baltimore, Maryland, USA, June 23-25 2014. c⃝2014 Association for Computational Linguistics A Step-wise Usage-based Method for Inducing Polysemy-aware Verb Classes Daisuke Kawahara† Daniel W. Peterson‡ Martha Palmer‡ †Kyoto University, Kyoto, Japan ‡University of Colorado at Boulder, Boulder, CO, USA dk@i.kyoto-u.ac.jp, {Daniel.W.Peterson, Martha.Palmer}@colorado.edu Abstract We present an unsupervised method for in- ducing verb classes from verb uses in giga- word corpora. Our method consists of two clustering steps: verb-specific seman- tic frames are first induced by clustering verb uses in a corpus and then verb classes are induced by clustering these frames. By taking this step-wise approach, we can not only generate verb classes based on a massive amount of verb uses in a scalable manner, but also deal with verb polysemy, which is bypassed by most of the previous studies on verb clustering. In our exper- iments, we acquire semantic frames and verb classes from two giga-word corpora, the larger comprising 20 billion words. The effectiveness of our approach is veri- fied through quantitative evaluations based on polysemy-aware gold-standard data. 1 Introduction A verb plays a primary role in conveying the meaning of a sentence. Capturing the sense of a verb is essential for natural language processing (NLP), and thus lexical resources for verbs play an important role in NLP. Verb classes are one such lexical resource. Manually-crafted verb classes have been devel- oped, such as Levin’s classes (Levin, 1993) and their extension, VerbNet (Kipper-Schuler, 2005), in which verbs are organized into classes on the basis of their syntactic and semantic behavior. Such verb classes have been used in many NLP ap- plications that need to consider semantics in par- ticular, such as word sense disambiguation (Dang, 2004), semantic parsing (Swier and Stevenson, 2005; Shi and Mihalcea, 2005) and discourse pars- ing (Subba and Di Eugenio, 2009). There have also been many attempts to auto- matically acquire verb classes with the goal of ei- ther adding frequency information to an existing resource or of inducing similar verb classes for other languages. Most of these approaches assume that all target verbs are monosemous (Stevenson and Joanis, 2003; Schulte im Walde, 2006; Joa- nis et al., 2008; Li and Brew, 2008; Sun et al., 2008; Sun and Korhonen, 2009; Vlachos et al., 2009; Parisien and Stevenson, 2010; Parisien and Stevenson, 2011; Falk et al., 2012; Lippincott et al., 2012; Reichart and Korhonen, 2013; Sun et al., 2013). This monosemous assumption, however, is not realistic because many frequent verbs actually have multiple senses. Moreover, to the best of our knowledge, none of the following approaches at- tempt to quantitatively evaluate soft clusterings of verb classes induced by polysemy-aware unsuper- vised approaches (Korhonen et al., 2003; Lapata and Brew, 2004; Li and Brew, 2007; Schulte im Walde et al., 2008). In this paper, we propose an unsupervised method for inducing verb classes that is aware of verb polysemy. Our method consists of two clustering steps: verb-specific semantic frames are first …

Content truncated to 500 words