Clustering polysemic subcategorization frame distributions semantically

Abstract

Previous research has demonstrated the utility of clustering in inducing semantic verb classes from undisambiguated corpus data. We describe a new approach which involves clustering subcategorization frame (SCF) distributions using the Information Bottleneck and nearest neighbour methods. In contrast to previous work, we particularly focus on clustering polysemic verbs. A novel evaluation scheme is proposed which accounts for the effect of polysemy on the clusters, offering us a good insight into the potential and limitations of semantically classifying undisambiguated SCF data.

Syntactic Transfer Patterns of German Particle Verbs and their Impact on Lexical Semantics

“…In this paper we investigate the influence of syntax, which represents one of the possible feature sources. Syn-tactic subcategorization frames tend to be good predictors for the semantics of verbs in general: verbs that are similar in meaning also tend to have similar subcategorization frames and selectional preferences (Schulte im Walde, 2000;Merlo and Stevenson, 2001;Korhonen et al, 2003;Schulte im Walde, 2006a;Joanis et al, 2008). But, as we will show below, PV-BV pairs tend to have a special behavior with respect to their subcategorization, even if their meanings are closely related.…”

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

“…First, we make multiple data points for each verb to deal with verb polysemy (cf. polysemy-aware previous studies still represented a verb as one data point (Korhonen et al, 2003;Miyao and Tsujii, 2009)). To do that, we induce verb-specific semantic frames by clustering verb uses.…”

“…The most closely related work to our polysemyaware task of unsupervised verb class induction is the work of Korhonen et al (2003), who used distributions of subcategorization frames to cluster verbs. They adopted the Nearest Neighbor (NN) and Information Bottleneck (IB) methods for clustering.…”

“…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 attempt to quantitatively evaluate soft clusterings of verb classes induced by polysemy-aware unsupervised approaches (Korhonen et al, 2003;Lapata and Brew, 2004;Li and Brew, 2007; Schulte im Walde et al, 2008).…”

Comparing and combining semantic verb classifications

“…As for the acquisition of verb classes, automatic methods such as those suggested by Korhonen et al (2003), Schulte im Walde (2006), or Joanis et al (2008) have the potential of reducing the prohibitive cost of manual methods. However, they require decisions about both the experiment setup (with regard to feature selection) and the choice of a manually constructed gold standard for evaluation.…”