Estudo prosódico das disfluências de reparo

• Ano de defesa: 2009
• Autor(a) principal: Alexandre Delfino Xavier
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; UFMG
• Palavras-chave em Português: ESTUDOS LINGÜÍSTICOS/M; Análise prosódica (Linguística); Lingüística; Fonética acústica; Distúrbios da fala; Entonação (Fonética); Lingua portuguesa Versificação
• Link de acesso: http://hdl.handle.net/1843/ALDR-7R6GLS

The aim of this dissertation is to describe the prosodic features present in spontaneous speech containing repair disfluências and compare which of these features can diferentiate from the fluent speech. Repair disfluencies strings of sentences in which the speaker corrects himself just after noticing the production of a wrong information. Although they are often considered only as an error during the linguistic processing, thedisfluencies can work as an evidence for the production and comprehension language systems. A major part of studies on disfluencies is concerned to aspects as syntax and semantics. The proposal of this work follows a group of investigations that analise theinfluence of phonological aspects of the disfluencies for the linguist processing models (Ferreira, 2005; Fodor, 1998, 2002; Levelt, 1989, Shriberg, 1994, 1999; among others). Starting from the self-monitoring model and from the remarks on the prosodic marking in repair disfluencies described by Levelt (1983, 1989), we evaluated acoustic features that may characterize a disfluent speech. We used as a basis for the repair corpus the larger previous corpus from Magalhães (2000) about the mineiro accent from Belo Horizonte. The first feature to be analised was the pitch contour, especially the part correspondentto the wrong information (reparandum) e to the corrected information (repair). Furthermore, we analised the behaviour of the F0 peak during the two parts. After that, we analised the issue on the picht contour interruption as an acoustic evidence for the end of the reparandum. The second feature was the intensity values; as the F0 analysis, we compared the value of intensity during the two parts. The third feature was the speech rate, comparing the elocution rate and the articulation rate of speech. We analised the occurence of silent pauses and filled pauses too. Last, we studied the presence of other features, such as lengthening, laryngelizations and creaky voice. Indeed, the results of the acoustic and the statistical analysis point towards to our original hypothesis that the speaker produces significant acoustic cues that may show the difference between the reparandum and the repair. Considering the F0 and the intensity, there is a contrastive accent marking, as a way of highlighting or emphatizing the repair information. The pitch contour interruption is seen as a result of the self-monitoring error detection. The lengthening and the pauses may occur when the speaker needs to plan more the upcoming message as he speaks.The results from this dissertation corroborate the ideas that prosody is an relevant component in the phrasal processing and that the speaker has a self-monitoring speech mecanism.