Descrição de expressões fixas do português brasileiro para processamento automático de linguagem natural

Abstract

This research proposes a study of fixed expressions of Portuguese in Brazil, especially those with names of body parts, identified in newspaper texts and dictionaries. The description and formalization of these fixed expressions are of utmost importance to Automatic Processing of Natural Language. We selected a corpus with 558 fixed expressions, 351 are of verbal argument structures and 207 nouns argument structures. These expressions are analyzed from formal linguistic criteria to identify the morphosyntactic-semantic behavior of each one and, from these structural properties observed in the context of use, we propose an encoding to formalize them, with the aim to be inserted into a computing environment, expanding the descriptive studies on the lexicon of Portuguese and correcting deficiencies that might impair the quality of a text processed automatically. The formal representation of these fixed expressions is essential so that we can include them in an electronic dictionary, and thus use them at Unitex software or any other morphosyntactic analyzer.