Navegando por Assunto "Brazilian portuguese"
Agora exibindo 1 - 3 de 3
- Resultados por página
- Opções de Ordenação
Item Acesso aberto (Open Access) Análise perceptual da harmonia vocálica na fala Belenense(Universidade Federal do Pará, 2024-08-29) XAVIER, Francisco Cavalcante; CRUZ, Regina Célia Fernandes; http://lattes.cnpq.br/3307472469778577The present work aims to investigate, perceptually, the productivity of vowel harmony (VH) triggered by low vowels in the variety of Portuguese spoken in Belém, PA (Brazil). To this goal, a corpus was formed with 42 paroxytone words, mostly in the CV'CV.CV syllabic mold, in which /e/ and /o/ alternate in the pretonic syllable and /i, e, ɛ, a, ᴐ, o, u/, in the tonic syllable. Using the Wideo Text-to-Speech Software converter, three variants were generated as sound stimuli for each item in the corpus, according to the height of the pretonic vowel: for , [i], [e ], [E]; for , [u], [o], [O]. A sample of 60 Belenenses (people from Belém), stratified by gender, age group and education, responded to a questionnaire implemented on the Gorilla Experiment BuilderTM platform, version 4, with two data collection protocols: I - Frequency Assessment (FA); II - Identification Assessment (IA). For the FA, the main protocol, participants attributed to the variants of , an approximate frequency of occurrence/use in Belenense speech, based on the following scalar indexes: Never, Rarely, Sometimes, Almost always. In the IA, the participants’ assessment was taken to identify the three variants as, in fact, different. For the statistical treatment of the 10,080 data collected, Simple Correlation and Binary Logistic Regression analyzes were applied, using the R Program, version 2024.04.1. Taking , as dependent variables and the seven stressed vowels as independent ones, the results revealed that, in Belém's speech, in general, based on the choice of the full occurrence index, Almost always: (a) the high variants are the least frequent – [i], with a relative frequency of .13 and [u], with .20; (b) the medium variants are largely predominant, adjusting relatively well to all stressed vowels – [e], .77 and [o], .75; (c) the low variants are in second place as most frequent – [E], .43 and [O], .41 –, but, strongly attracted by low stressed vowels, they take the hegemony of the mids in this structural context – [E] , .82; [O], .78. Finally, for the external factors: (a) older speakers highlighted the hegemony of VH – [E], .83; [O], .83; (b) younger speakers attenuated it – [E], .72; [O], .65; c) individuals trained in the research area had higher discrimination rates between the three variants in IA – , .89; , .87. In this view, in the perceptual field, a phonological rule of HV triggered by low vowels was attested in full operation in Belém speech, as acoustic studies had already pointed out (Sousa, 2010; Fagundes, 2015; Souza, 2020). The phenomenon proved to be productive with all low vowels in stressed syllables, both on and , contradicting, at this specific point, the acoustic signal (Souza, 2020).Item Acesso aberto (Open Access) Desenvolvimento de recursos para a construção de um sistema texto-fala para o português brasileiro(Universidade Federal do Pará, 2010-12-23) COUTO, Igor Costa do; KLAUTAU JÚNIOR, Aldebaro Barreto da Rocha; http://lattes.cnpq.br/1596629769697284Text-to-speech (TTS) is currently a mature technology that is used in many applications. Some modules of a TTS depend on the language and, while there are many public resources for English, the resources for some underrepresented languages are still limited. This work describes the development of a complete TTS system for Brazilian Portuguese (BP) which expands the already available resources. The system uses the MARY framework and is based on the hidden Markov model (HMM) speech synthesis approach. Some of the contributions of this work consist in implementing syllabification, determination of stressed syllable and grapheme-tophoneme (G2P) conversion. This work also describes the steps for organizing the developed resources and implementing a BP voice within the MARY. These resources are made available and facilitate the research in text normalization and HMM-based synthesis for BP.Item Acesso aberto (Open Access) Padrão geoprosódico do português falado na Amazônia Oriental: análise comparativa de Belém, Macapá e São Luís(Universidade Federal do Pará, 2021-05-19) LEMOS, Rosinele Lemos e; CRUZ, Regina Célia Fernandes; http://lattes.cnpq.br/3307472469778577The main objective of this thesis is to carry out a geoprosodic mapping of three inquiry points of Brazilian Portuguese (PB) spoken in the Eastern Amazon: Belém (PA), Macapá (AP) and São Luís (MA), using modal intonation as the object of analysis. It is a study based on the assumptions of Sociophonetics (THOMAS, 2011) for understanding a speech acoustic analysis that considers the performance of social factors such as sex and schooling in describing the phenomenon analyzed. The interdialetal comparison of the three capitals comprises an acoustic analysis of the physical parameters of fundamental frequency (in st), duration (in Z-score) and intensity (in Z-score), observed in neutral declarative and interrogative sentences. For data collection, the methodology established by the AMPER project was used. The treatment of the data consisted of several stages, two of which are in line with the AMPER methodology: a) data coding; b) isolation of audios in individual files. The other steps were carried out with the support of scripts created by Albert Rilliard (LIMSI-CNRS), used in the automatic segmentation of the data, namely: c) script lance_batch_easyalign_v3.praat to obtain the textgrid of the .wav files; d) script of correction_segmentation.praat for phonetic segmentation in the praat program 6.0.39, e) script AMPER_Textgrid2Txt_V3_boucle_DepoisEasyAlign_v2.praat that generated the files with the acoustic parameters; f) normalization of acoustic parameters. With the treated data, the acoustic measurements of the parameters of fundamental frequency (f0), duration and intensity were obtained, which were organized in an Excel spreadsheet for later statistical treatment and graphing in software R. In order to validate the acoustic analyzes, twonon-parametric statistical tests were applied (Kruskal-Wallis and Wilcoxon) and the Post hoc test. For this research, speech samples were collected from eighteen speakers, stratified by sex, education (elementary, high school and university) and place. The total corpus consisted of 33 neutral declarative sentences and 33 total interrogatives, and the words used to compose the sentences contemplated the three accentuated staves of the Portuguese (stress on the last middle and first syllable and the extension of the sentences contains 10, 13 or 14 vowels, with sentences repeated six times by each speaker. In total there were 3.564 data analyzed (33 sentences x 2 categories x 3 best repetitions x 6 announcers x 3 places). This demonstrates that there are significant differences both for the place factor and for the social factors (sex and education of the speakers). The confronted capitals registered a similar prosodic pattern with respect to the variations of f0 in the acoustic characterization of the neutral declarative and total interrogative sentences. The results showed that the paraense, macapaense and ludovicense capitals followed the intonation pattern described for PB, with an elevation of the intonation peak in the pretonic and a drop of f0 in the stressed syllable in the neutral declarative modality (SILVESTRE, 2012). The total interrogatives showed lower f0 values in the pretonic ones with an increase in the final tonic, followed by a drop in the postonic ones when they existed, the circumflex pattern described by Moraes (1993) for PB. Regarding the intonational contour, the three capitals of the Eastern Amazon presented a similar f0 contour in the final noun phrase (SNF). On the other hand, there was a difference in the height of f0, since Macapá presented higher contours of f0 in the final stressed syllables, in the three accent staves, for the two phrasal modalities; while São Luís and Belém registered height of f0 approximate averages. The results of the acoustic analysis of the schooling factor showed that Macapá presented a greater variation of f0 in the three levels of education, in the three accentual guidelines of Portuguese for the two phrasal modalities. The duration parameter confirmed higher values in the tonic, with a similar production time in the two phrasal modalities, in the three mapped capitals, which is not a peculiarity of the modal intonation, but which is linked to the lexical accent. Variations in the intensity parameter are related to modal intonation. Therefore, the prosodic pattern of the target places found in this study behaves similarly to PB, since there was no difference in terms of modal intonation between the pattern verified in PB and the Portuguese spoken in the three capitals of the Eastern Amazon.