Unsupervised Extraction of Syntactic Patterns for Identifying Lexical Opposition Relations in Spanish

Abstract

Lexical resources are difficult, costly and time-consuming to produce and maintain. In this paper, we propose a methodology for the automatic extraction of lexical opposition relations, in particular antonyms and co-hyponyms, by means of juxtaposed syntactic patterns, in order to contribute to the construction and expansion of lexical resources. Using an extensive Spanish text corpus, specific rules and textual analysis tools developed in Python, a methodology based on three modules was implemented: (1) extraction of repetition patterns, (2) filtering of co-hyponyms by symmetry, and (3) binarization of multiple candidate pairs of antonyms. The system showed a high level of precision in classifying cohyponyms and antonyms, according to an evaluation based on human annotations. The results demonstrate that the methodology is able to identify oppositional lexical relations efficiently, without relying on external linguistic resources, which can contribute to the automated enrichment and maintenance of lexical resources.