Is there content in empty heads?

Abstract

We describe a technique for automatically constructing a taxonomy of word senses from a machine readable dictionary. Previous taxonomies developed from dictionaries have two properties in common. First, they are based on a somewhat loosely defined notion of the IS-A relation. Second, they require human intervention to identify the sense of the genus term being used. We believe that for taxonomies of this type to serve a useful role in subsequent natural language proce,sing tasks, the taxonomy must be based on a consistent use of the IS-A relation which allows inheritance and transitivity. We show that hierarchies of this type can be automatical!y constructed, by using the semantic category codes and the subject codes of the Longman Dictionary of Contemporary English (LDOCE) to disambiguate the genus terms in noun definitkms. In addition, we discuss how certain genus terms give rise to other semantic relations between definitions.

Natural Language Processing: The PLNLP Approach

“…Wl$l stands for Webster, homograph 1, sense 1. '0' is assigned if there is only one (unnumbered) homograph or sense in However, although Brachman 1983 discusses the problem of different kinds of IS-A relations in the construction of knowledge bases, little attention has been paid to the complexity of IS-A relations in the construction of taxonomies from dictionaries (the exceptions being Guthrie et al 1990, andVeronis andIde 1990). The two types of IS-A relation that most frequently hold between headwords and their genus terms in dictionary definitions are the IS-A type/subset relation, which Brachman does discuss, and the IS-A synonym relation, which he does not.…”

“…Although many systems have undertaken one or two of these tasks, only Guthrie et al have tried all three (Guthrie et al 1990). Chodorow et al 1985 posited a category "empty head" for genus terms which were not in an IS-A relation with the headword.…”

“…These words triggered the search for additional heads in the definition. Guthrie et al 1990, in developing genus disambiguation procedures based on the semantic codes in the Longman Dictionary of Contemporary English (LDOCE), reduced the set of "truly empty" heads to four: one, any, kind, and type. However, they also observed that other syntactic heads such as collectives (set, group, collection) do in fact provide important semantic information.…”

Acquisition of OWL DL Axioms from Lexical Resources

“…In particular, the rules relying on Hearst-style patterns [5] for the identification of hyponymy relationships may be mislead by expressions such as one, any, kind, type. This phenomenon has already been described [9,10] and could be handled by appropriate exception rules. An alternative solution to this and similar problems could be to increase the expressiveness of the rule language used in the transformation process.…”

“…In the medium term, we expect to develop a corresponding system as part of a powerful ontology engineering environment like OntoStudio, Protégé, or the forthcoming NeOn Toolkit 9 . It will also be worth investigating the use of LExO for automated question answering.…”