Estudo comparativo entre análise multidimensional lexical e modelagem de tópicos

• Autor(a) principal: Owa, Denis Luiz Marcello
• Data de Publicação: 2021
• Tipo de documento: Tese
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da PUC_SP
• Texto Completo: https://repositorio.pucsp.br/jspui/handle/handle/24252
• topic: CNPQ::LINGUISTICA, LETRAS E ARTES::LINGUISTICA::LINGUISTICA APLICADA; Linguística de corpus; Análise multidimensional lexical; Modelagem de tópicos; Análise Multidimensional Lexical; Lexical multidimensional analysis; Topic modeling

The main objective of this work is to compare two theme / topic / discourse modeling methods from a corpus perspective, namely, Lexical Multidimensional Analysis (LMDA) and Topic Modeling (TM). The comparison is based on a range of different corpora, namely: a corpus consisting of 1,000 academic articles written in English (CAPO - Corpus of Articles from PLOS ONE), three corpora in English (the Sally Mann Corpus; the Children Culture Corpus – CHICC; and the Corpus of Religious and Dogmatic English Language Texts - CReDELT), in addition to a corpus of Portuguese (Corpus of Ethnic-Racial Records - CRER). As mentioned, we used Lexical Multidimensional Analysis as a means to detect the major dimensions of lexical variation, identified through the interpretation of the semantic fields underlying the co-occurrence of the most salient lexis. This framework was developed by Berber Sardinha (2016, 2017, 2019, 2021) based on the framework developed by Biber in the 1980s (Biber, 1988) for the analysis of register variation. In addition to Lexical MD Analysis, in this project we used Topic Modeling (TM), which consists of identifying sets of words commonly occurring across texts as a means to detect the major ‘topics’ in a text collection. The central works of the area, with relevance for this study are Blei; NG; Jordan (2003) and Srinivasa-Desikan (2018). Although previous research exists for Lexical MD Analysis (e.g. Berber Sardinha (2017, 2021) and for Topic Modeling (e.g. Griffiths and Mark Steyvers, 2004; Wang and Blei, 2011 Younus et al., 2014; Murakami et al., 201; and Krasnov, 2018), no previous research exists that compares and contrasts the results of these two frameworks, despite the fact that both are used for similar purposes in the literature. Thus, this dissertation seeks to make an original contribution by comparing and contrasting these two methods