Automated Induction of Sense in Context

Automated Induction of Sense in Context James PUSTEJOVSKY, Patrick HANKS, Anna RUMSHISKY Brandeis University Waltham, MA 02454, USA {jamesp,patrick,arum}@cs.brandeis.edu Abstract In this paper, we introduce a model for sense as- signment which relies on assigning senses to the contexts within which words appear, rather than to the words themselves. We argue that word senses as such are not directly encoded in the lexicon of the language. Rather, each word is associated with one or more stereotypical syntagmatic pat- terns, which we call selection contexts. Each selec- tion context is associated with a meaning, which can be expressed in any of various formal or com- putational manifestations. We present a formalism for encoding contexts that help to determine the semantic contribution of a word in an utterance. Further, we develop a methodology through which such stereotypical contexts for words and phrases can be identified from very large corpora, and sub- sequently structured in a selection context dictio- nary, encoding both stereotypical syntactic and se- mantic information. We present some preliminary results. 1 Introduction This paper describes a new model for the acquisi- tion and exploitation of selectional preferences for predicates from natural language corpora. Our goal is to apply this model in order to construct a dic- tionary of normal selection contexts for natural lan- guage; that is, a computational lexical database of rich selectional contexts, associated with procedures for assigning interpretations on a probabilistic basis to less normal contexts. Such a semi-automatically developed resource promises to have applications for a number of NLP tasks, including word-sense disam- biguation, selectional preference acquisition, as well as anaphora resolution and inference in specialized domains. We apply this methodology to a selected set of verbs, including a subset of the verbs in the Senseval 3 word sense discrimination task and report our initial results. 1.1 Selectional Preference Acquisition: Current State of the Art Predicate subcategorization information constitutes an essential part of the computational lexicon entry. In recent years, a number of approaches have been proposed for dealing computationally with selec- tional preference acquisition (Resnik (1996); Briscoe and Carroll (1997); McCarthy (1997); Rooth et al. (1999); Abney and Light (1999); Ciaramita and Johnson (2000); Korhonen (2002)). The currently available best algorithms developed for the acquisition of selectional preferences for pred- icates are induction algorithms modeling selectional behavior as a distribution over words (cf. Abney and Light (1999)). Semantic classes assigned to predi- cate arguments in subcategorization frames are ei- ther derived automatically through statistical clus- tering techniques (Rooth et al. (1999), Light and Greiff (2002)) or assigned using hand-constructed lexical taxonomies such as the WordNet hierarchy or LDOCE semantic classes. Overwhelmingly, Word- Net is chosen as the default resource for dealing with the sparse data problem (Resnik (1996); Abney and Light (1999); Ciaramita and Johnson (2000); Agirre and Martinez (2001); Clark and Weir (2001); Carroll and McCarthy (2000); Korhonen and Preiss (2003)). Much of the work on inducing selectional prefer- ences for verbs from corpora deals with predicates in- discriminately, assuming no differentiation between predicate …

Content truncated to 500 words