The Formalization of WordNet by Methods of Relational Concept Analysis
Uta Priss · The MIT Press eBooks · 1998
this paper was partially supported by the Zentrum fur Interdisziplinare Technikforschung, Darmstadt be described by their hypernyms, attributes, and other relations, but are not denominated by words. Two formal contexts are needed for the study of semantic relations, a denotative context, which contains the denotational meanings of word forms (denotata) and their conceptual ordering, and a lexical context, which has the words as constitutive elements. The words are always assumed to be disambiguated (for example by WordNet sense numbers) to avoid problems of polysemy and homonymy. A denotative context is usually incomplete because it is not possible to write a list of all denotata of a language. But, as semantic relations refer to relations among denotata, they cannot be defined on words without studying the denotata in a denotative context. Examples for lexical contexts are lexical fields (Kipke & Wille, 1987). Furthermore, every dictionary or thesaurus can be interpreted as a lexical context. Words are names for concepts in a denotative context and formal objects in a lexical context. Therefore it has to be investigated whether semantic relations have the same properties in both contexts. WordNet is formalized as such a lexical context, but only the noun synsets are investigated in this chapter (Section 7.3). The other parts of speech and more details can be found elsewhere (Priss, 1996). In Section 7.7, three examples of the meronymy relation in WordNet show how this theoretical framework can be used to find irregularities among the semantic relations in WordNet1.5