Knowledge-based WSD

Work on WSD reached a turning point in the 1980s and 1990s when large-scale lexical resources such as dictionaries, thesauri, and corpora became widely available. The work done earlier on WSD was theoretically interesting but practical only in extremely limited domains. Since Lesk1986, many researchers have used machine-readable dictionaries (MRDs) as a structured source of lexical knowledge to deal with WSD. These approaches, by exploiting the knowledge contained in the dictionaries, mainly seek to avoid the need for large amounts of training material. Agirre2001tsd distinguish ten different types of information that can be useful for WSD. Most of them can be located in MRDs, and include part of speech, semantic word associations, syntactic cues, selectional preferences, and frequency of senses, among others.

In general, WSD techniques using pre-existing structured lexical knowledge resources differ in:

Lesk1986 proposes a method for guessing the correct word sense by counting word overlaps between dictionary definitions of the words in the context of the ambiguous word. Cowie1992 uses the simulated annealing technique for overcoming the combinatorial explosion of the Lesk method. Wilks1993 use co-occurrence data extracted from an MRD to construct word-context vectors, and thus word-sense vectors, to perform a large set of experiments to test relatedness functions between words and vector-similarity functions.

Other approaches measure the relatedness between words, taking as a reference a structured semantic net. Thus, Sussna1993 employs the notion of conceptual distance between network nodes in order to improve precision during document indexing. Agirre1996 present a method for the resolution of the lexical ambiguity of nouns using the WordNet noun taxonomy and the notion of conceptual density. Rigau1997 combine a set of knowledge-based algorithms to accurately disambiguate definitions of MRDs. Mihalcea1999 suggest a method that attempts to disambiguate all the nouns, verbs, adverbs, and adjectives in a given text by referring to the senses provided by WordNet. Magnini2002 explore the role of domain information in WSD using WordNet domains Magnini2000; in this case, the underlying hypothesis is that information provided by domain labels offers a natural way to establish semantic relations among word senses, which can be profitably used during the disambiguation process.

Although knowledge-based systems have been proven to be ready-to-use and scalable tools for all-words WSD because they do not require sense-annotated data Montoyo2001, in general, supervised, corpus-based algorithms have obtained better precision than knowledge-based ones.