Hierarchical and distributional lexical field theory: A critical and empirical development of louw and nida's semantic domain model

Abstract

This paper argues that the underdeveloped notion of semantic similarity in Louw and Nida's lexicon can be improved by taking account of distributional information. Their use of componential analysis relies on a set of metalinguistic terms, or components, that are ultimately arbitrary. Furthermore, both the polysemy within their semantic domains and the organization of those domains problematize their categories. By contrast, distributional data provide an empirical measurement of semantic similarity, and lexicogrammatical categorization provides a non-intuition-driven principle of classification. Distributional data is gathered by word embedding, and lexicogrammatical categorization is based largely on a derived metric of abstraction. This argument is tested by considering probable semantic field relationships for a number of Greek lexemes. Ultimately, this approach provides directions to address some of the critical weaknesses in semantic domain or semantic field theory as applied to the study of Hellenistic Greek, by introducing empirical means of approximating lexical fields. VC 2018 Oxford University Press. All rights reserved.