Marcadores discursivos em uma perspectiva informal: análise prosódica e estatística

• Ano de defesa: 2019
• Autor(a) principal: Oliver Renato de Araujo Gobbo
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; UFMG
• Palavras-chave em Português: linguística de corpus; linguística quantitativa; prosódia; marcadores discursivos; teoria da Lingua em Ato; Análise prosódica (Linguística); Marcadores discursivos; Linguística de corpus
• Link de acesso: http://hdl.handle.net/1843/LETR-BAVN6N

Discourse markers (DMs) are still a very problematic linguistic category, despite many dedicated studies over the past four decades. There is no consensus in the literature on terminology, on the category definition and on how to identify the DMs and their functions. A proposal for the study of DMs is presented by the Language into Act Theory (LAcT) (CRESTI, 2000; MONEGLIA, 2005; RASO, 2012; MONEGLIA and RASO, 2014), elaborated through spontaneous speech corpora observation. L-AcT supports that communicative functions are carried out through prosody, not through the lexical item, as it is traditionally adopted in DMs studies. From thisperspective, DMs would be informational units of dialogic type (DUs) and not textual type, as they address the interlocutor performing several interactional functions. DMs are also always isolated in intonational units and they do not compose the utterance propositional content. The present work investigates DMs in the informal section of C-ORAL-BRASIL spontaneous speech subcorpus (Mini BR Informal). In a first step, the Mini BR Informal was reviewed in a search for DUs presenting prosodic form matching three types already satisfactorily described in the literature (RASO, 2014; RASO and VIEIRA, 2016): conative, allocutive and incipt. These DUs were submitted to linear discriminant analysis (LDA) intended to investigate statistically which prosodic features better differentiate the types. The statistical model achieved 84.6% classification accuracy and pointed to four mostly discriminating features: average f0; average intensity; initial f0 slope and maximum f0 position. Other units, which do not match the three types of DMs already described, were submitted to cluster analysis with the aim of statistically determining patterns based on their prosodic features. Results identified three major groups that merged by the similarity of the investigated features, proposing the hypothesis these groups relate to three other DUs types.