ARTIFICIAL INTELLIGENCE IN THE CONSTRUCTION OF PHRASEOLOGICAL CORPORA: A STRUCTURED REVIEW OF EXTRACTION, DISAMBIGUATION, AND VALIDATION METHODS

Abstract

Phraseological units – idioms, set expressions, light-verb constructions, and other non-compositional word combinations – are among the most difficult objects a corpus can be built around. They are semantically opaque, formally variable, culturally specific, and individually rare, which means that neither dictionary compilation nor manual corpus annotation scales to the volume required for quantitative study. Artificial intelligence (AI) has been applied to every step of this problem, from candidate generation to figurative-literal disambiguation, and a substantial body of resources and shared tasks now exists. That body of work has developed largely inside computational linguistics and has not been consolidated for the phraseological research community.