Argument-structure and implicational constructions in a knowledge base
DOI:
https://doi.org/10.7764/onomazein.35.04Keywords:
Constructions, Construction Grammar, cognitive models, knowledge-base, Lexical Constructional Model, Natural Language Processing.Abstract
This paper provides formalized, machine-tractable representations of two broad kinds of constructional configuration, argument-structure and implicational constructions, on the basis of previous linguistic analyses. It discusses computational implementation requirements on constructional description. In this respect, the paper argues that the Goldbergian approach (cf. Goldberg, 2006) provides the best fit for the implementation of implicational constructions, while a “mini-constructionist” account (cf. Boas, 2014) is suitable for argumentstructure constructions. Because of their representativeness, we have chosen to illustrate our discussion by making reference to the family of English resultatives, which are argumentstructure constructions, and to the family of Wh-attitudinal constructions, which are implicational. Computational implementation demands that the members of the family of the resultative be split into mini-constructions, while the complexity of implicational configurations requires that different formal variants be grouped together under one single computational representation. The paper further makes explicit proposals for the machine tractability of lexical-constructional integration and of meaning implications that have reached constructional status through entrenchment, two problems that remain unsolved within standard computational approaches to language processing.