Não é nada não: uma análise das lexias negativas do português para anotação nas Universal Dependencies

Abstract

Negation is a sine qua non of natural languages, there is no language that does not present some way to express it in its system (HORN and WANSING, 2020). Despite being very productive and indispensable for natural languages, studies in natural language processing that involve negation is extremely recent (JIMÉNEZ-ZAFRA et al., 2020), which makes the material on this subject scarce. Therefore, this work presents an analysis of the Portuguese negative clues (não, nunca, jamais, sequer, tampouco, nada, ninguém, nenhum, sem, nem, exclusive, excluso, exceto, fora, afora, salvo, menos, senão e tirante) using the Universal Dependencies (UNIVERSAL DEPENDENCIES, 2021) methodology to increase the amount of information available of each element. The analysis consisted of a literature review of each clue and a search for them in the corpora UD Portuguese Bosque (RADEMAKER et al., 2017) and Porttinari-base (LOPES et al., 2022a) and the lexicon/lexical tool PortiLexicon-UD (LOPES et al., 2022b) as well. With the results, the morphological features and some syntactic relationships of the clues were analyzed, proposing possible corrections and optimizations in the current annotation of each corpus as proposals to improve the lexical tool. The review of negative clues is extremely important for future works that intend to perform automatic linguistic analyzes that have the detection of negation, both of scope and focus, as part of their task.