Aprendizagem de categorias de palavras por análise distribucional resultados adicionais para Português Brasileiro

Abstract

A model of part-of-speech (or syntactic category) learning through distributional analysis – as a task in the language acquisition process – is presented here. It is based on Redington et al.'s (1998) model, but the input data used comes from publicly available corpora of both child-directed speech and speech between adults in Brazilian Portuguese. Results from four (out of nine) experiments are presented and discussed: experiments 2, 3, 4, and 7, of the original study. These experiments investigate variables such as the number of target and context words (2) and corpus size (4), and evaluate also the value of distributional information for different categories (3) and the learner's performance when functional categories are removed from the corpus (7). In general, our results support Redington et al.'s, although we find some few differences. We also evaluate the cosine metric, comparing it with performance obtained with the spearman rank correlation metric used in Redington et al.'s study. The latter seems to produce better results.

A informatividade da morfologia para a categorização distribucional de palavras

“…No capítulo 4, discutimos os resultados obtidos, partindo dos resultados quantitativos das sete simulações realizadas, com o intuito de verificar o desempenho geral do modelo, comparando com as simulações trazidas por Faria e Ohashi (2018), Faria (2019aFaria ( , 2019b). E, num segundo momento, apresentamos a análise qualitativa de três simulações, dando foco para aquelas que utilizam as informações morfológicas.…”