next up previous
Next: PARTOF relations Up: Lexical properties and relations Previous: Lexical properties and relations

ISA relations

The ISA relation is perhaps the fundamental lexical relation. The term is rather general, and covers relations which have been referred to in other formalisms and theoretical frameworks with terms such as: paradigmatic relation, classification, taxonomy, field, family, similarity, set partition, subset-set inclusion, element-set membership, generalisation, property, implication, inheritance.

Typical ISA relations define, in phonology, the natural classes characterised by distinctive feature vectors or by distributional classes based on syllable or word positions; in morphology, affix and stem classes; in phrasal syntax, parts of speech and constituent categories; in semantics, synonym, antonym and hyponym sets, or semantic fields. Closely related to the ISA relation is, in conventional artificial intelligence parlance, the HASPROP relation, which assigns the shared properties which characterise sets of objects which enter into ISA relations. In lexicology and lexicography, properties are generally partitioned into mutually non-coocurring groups, as values of attributes such as [CATEGORY: noun], where CATEGORY is an attribute with values which are members of a set of mutually exclusive elements (e.g. the traditional {noun, adjective, verb, adverb, article, pronoun, preposition, conjunction, interjection}).

In contemporary lexicographic methodology, inheritance lexica utilise ISA relations in order to generalise over properties shared by lexical entries, and generate very compact lexical representations in which many generalisable microstructure properties are inherited from higher level classes (cf. Section 3.4). In conventional lexica, ISA relations underlie synonym sets and cross references (`q.v.', `cf.', `cp.', `tex2html_wrap_inline1263') between entries.



Dafydd Gibbon
Thu Nov 19 10:12:05 MET 1998