Portuguese Speech Research Articles

A Small Brazilian Portuguese Speech Corpus for Speaker Recognition Study

A small Brazilian speech corpus was created for educational purposes to study a state-of-the-art speaker recognition system. The system uses the Gaussian Mixture Model (GMM) as a statistical model for speakers and employs the Mel-frequency cepstral coefficients (MFCC) as acoustic features. The results using clean and noisy speech are compatible with the expected results, showing that the bigger the mismatch between training and test conditions, the worse the results. The results also improve with the increase in the utterance length. Finally, the obtained results can be used as baselines to compare with other speaker statistical models created with different acoustic features in different acoustic conditions.

A Data-Centric Approach for Portuguese Speech Recognition: Language Model And Its Implications

Recent advances in Automatic Speech Recognition have made it possible to achieve a quality never seen before in the literature, both for languages with abundant data, such as English, which has a large number of studies and for the Portuguese language, which has a more limited amount of resources and studies. The most recent advances address speech recognition problems with Transformers based models, which have the capability to perform the speech recognition task directly from the raw signal, without the need for manual feature extraction. Some studies have already shown that it is possible to further improve the quality of the transcription of these models using language models within the decoding stage, however, the real impact of such language models is still not clear, especially for the Brazilian Portuguese scenario. Also, it is known that the quality of the data used for training the models is of paramount importance, however, there are few works in the literature addressing this issue. This work explores the impact of language models applied to Portuguese speech recognition both in terms of data quality and computational performance, with a data-centric approach. We propose an approach to measure similarity between datasets and, thus, assist in decision-making during training. The approach indicates paths for the advancement of the state-of-the-art aiming at Portuguese speech recognition, showing that it is possible to reduce the size of the language model by 80% and still achieve error rates around 7.17% for the Common Voice dataset. The source code is available at https://github.com/joaoalvarenga/language-model-evaluation.

Syllable position effects in the perception of L2 Portuguese /l/ and /ɾ/ by L1-Mandarin learners

This study reports syllable position effects on second language (L2) Portuguese speech perception, revealing that L2 segmental learning may be prone to an influence from the suprasegmental level. The results show that first language (L1) Mandarin learners had diminished performance on the discrimination between the target Portuguese liquids (/l/ and /ɾ/) and their position-dependent deviant productions, suggesting that the cause of their perceptual confusability differs across syllable positions. Another syllabic position effect was attested in the acquisition order (/l/onset > /l/coda, /ɾ/coda > /ɾ/onset), demonstrating that an L2 sound is not mastered equally in all positions. Furthermore, we also observed that an increase in L2 experience affected only the perceptual identification accuracy of [l], but not of [ɾ]. This seems to suggest that L2 experience may exert different degrees of impact, depending on the L2 segments. Both theoretical and methodological implications of these results are discussed.

Speech and Language Therapy Degree: reflections about school admission criteria in Portugal and worldwide

The purpose of the study is to identify the admission processes to the speech and language therapy degree worldwide and characterize how Portuguese speech and language therapists assess the candidates to the speech and language therapy degree. This is a descriptive, exploratory, and cross-sectional study. This study is divided into two studies. Questionnaires were used to collect data in both studies. The data were inserted and analyzed through descriptive statistics (via Statistical Package for the Social Sciences (SPSS) version 27) and content analysis. Considering study 1, the results show that there are no school admission criteria performed by a speech and language therapist in most countries. However, in some institutions, there are some specific procedures considering the acceptance of applicants to enter the speech and language therapy degree. Study 2 reveals that Portuguese speech and language therapists use mainly informal assessment procedures when performing school admission criteria. The skills/areas most valued as school admission criteria for accessing the degree are writing and reading skills. Most participants consider it essential to perform the school admission criteria before candidates can access the speech and language therapy degree. 
 There are different procedures worldwide considering the entrance in speech and language therapy degree. If school admission criteria should or should not exist is a matter of debate concerning inclusion and non-discrimination policies. Even in Portugal, where school admission criteria are mandatory, there is a high variability of procedures. All these results state the importance of discussing this matter in more depth, involving all the institutions and entities that formally represent the scientific area of speech and language therapy.

O efeito de tarefa na produção do ritmo por estudantes cantonenses de português como L2

ABSTRACT This study examines L2 Portuguese speech produced by eight native Cantonese speakers from Macao, China. The aims of this study are to investigate (1) whether the speech rhythm in L2 Portuguese is more source-like (more similar to Cantonese) or more target-like (more similar to Portuguese), and (2) whether L2 speech rhythm differs across three different tasks: a reading task, a retelling task, and an interpreting task. Seven rhythm metrics, i.e., %V, ΔC, ΔV, VarcoC, VarcoV, rPVI_C, and nPVI_V, were adopted for comparison and investigation. The results showed that L2 Portuguese rhythm produced by Cantonese speakers differed from L1 Portuguese speakers’ rhythm. R-deletion and vowel epenthesis were the reasons for the variabilities and instabilities of L2 Portuguese production by Cantonese learners, as they affect the duration and the number of vowel intervals and consonantal intervals. Moreover, in Cantonese learners’ L2 Portuguese production, the semi-spontaneous tasks (retelling and interpreting) presented a significant difference from the reading task. The driving force for such a difference was the cognitive load behind the tasks.

Neuromodulation: A combined-therapy protocol for speech rehabilitation in a child with cerebral palsy

IntroductionTranscranial direct current stimulation (tDCS) modulates cortical activity and potentiates functional gains achieved during therapeutic protocols. The aim of Integrative Speech Therapy Protocol is to rehabilitate speech in patients with impairments regarding neuropyschomotor development by combining oral motor stimuli, specific articulatory production, and the stimulation of phonological aspects of language. ObjectiveInvestigate the effect of transcranial direct current stimulation combined to integrative speech therapy in a child with cerebral palsy. MethodsWe performed a case study with tDCS and speech therapy in a patient with cerebral palsy and apraxia of speech. To assess the patient's speech, we used a parameterized test for the Brazilian Portuguese speech - ABFW. The CFCS and Vicking Speech Scales presented level IV and III, respectively. The patient underwent two periods with ten stimulation sessions each: first with anodal stimulation over Broca's area; and second over the left dorsolateral prefrontal cortex. Two indices were calculated: the percentage of consonants correct; and percentage of consonants correct-revised. Descriptive statistics were employed for the clinical data. For the outcomes, changes in each score were calculated as the difference in pre-intervention and post-intervention using Wilcoxon-Mann-Whitney test. ResultsIncreases were found in percentage of correct consonants indices as well as to produce two-syllable and three-syllable words after both types of stimulation, characterized mainly by correct vowels that marked the presence of the syllable. Number of phonemes increased 0 to 4 at first the stimulation and 4 to 10 at the second. ConclusionThe combined-therapy program contributed to improve the speech rehabilitation results in a patient with cerebral palsy.

Spontaneous emotional speech in European Portuguese using feeltrace system – first approaches

The present work studies spontaneous emotional speech in European Portuguese, comparing the values obtained in measuring acoustic parameters in this study with the ones known for other languages and for simulated. The spontaneous expression of emotions recorded in the street or from television programs may bring very relevant results, but even in those situations it is difficult to determine the individual’s true emotion (it is often only possible to be identify through the context) (Patrick, 2008). The main focus of our study is on voice related parameters such as F0, jitter, shimmer and HNR that will be liable to compare with previous studies of emotions produced by an actor. Using the Feeltrace system it is possible to better describe the sensation caused by the stimuli (in any of the emotions in question) and describe in more dynamic manner the emotion. We also realize that, although the emotions produced by actors show some differences, because they are necessarily exaggerated, the parameter values are all very similar, valuing the studies that are usually obtained based on utterances produced by performers. Given the scarcity of studies on voice quality in European Portuguese, it is important to highlight that this work presents an original corpus specifically created for this study. We recognize the limitations of the presented research seeing the results as first approaches that must be deeply studied being important to areas such as robotics and health technology.

OKIoT: Trade off analysis of smart speaker architecture on open knowledge IoT project

Successful deployment of smart environments based on Internet of Things (IoT) technologies face the challenges of constrained devices. Processing, bandwidth, and memory limitations must be considered when designing IoT systems to meet non-functional requirements. Software Engineering methods such as trade-off analysis could enhance the quality of IoT architectures at the design phase and bring benefits to end-users. These methods were applied to IoT projects specification during a short-term course (2019). Smart home capstone projects (2017, 2019) built with the proposed rationale are available on a public repository. Basic smart speaker services API supported qualitative and quantitative analysis of Brazilian Portuguese speech audios, smart speaker architecture Petri Net and queue models simulations on PIPE, and JMT tools enabled architecture trade-off analysis. Security vs. performance and accuracy vs. response time results provided insights for an optimized smart speaker open architecture.

Análise de alguns traços fonológicos graduais na escrita sob a perspectiva da sociolinguística educacional

The Brazilian school faces challenges, especially with regard to writing or, more specifically, to learning the standard norm that is intensively disseminated in school textbooks. What is ignored in the teaching / learning process of the written modality is the influence of orality, how speech habits are frequently represented in school texts and how they reflect variable linguistic traits common in Brazilian Portuguese (PB) speech. Thus, this article aims to discuss gradual phonological traits in writing present in 6th grade school essays according to the principles of Educational Sociolinguistics (BORTONI-RICARDO, 2005). The three most frequent deviations according to the research by Sene (2018) were selected and analyzed, through which it was observed that the principles of Educational Sociolinguistics are crucial to understand that such deviations should not be seen as a source of conflict in the school space, but as an opportunity for reflection on the importance of linguistic variation and sociolinguistic knowledge.

Globalisation, cultural diversity and Lusophony: trans-spatial circulation of Portuguese speech and its relationship with other speeches

I shall discuss several questions involving the concepts of globalisation, cultural diversity and lusophony in a transdisciplinary fashion, resorting to the notions of crossings and passages, mediated by Applied Linguistics and supported by Language Planning and Language Policy. The crossing, wrapped in novelty and perilousness, speaks of oceans, affording us complex challenges and, at times, frightening us. The passage seems to be more controlled, and associated with less enigmatic and more secure experiences, perhaps because both banks of a river can usually be seen from any point as we pass through it. In a river, we travel across from one bank to the other. And in the ocean, what happens? Lusophony and its contextualisation, in my opinion, is the crossing, viewed from a transdisciplinary perspective and through applied linguistics, including cultural diversity in a world claimed as global. Through Language Planning and Language Policy, both established as useful tools for a possible outline of some futuristic notion, Lusophony is here understood as passage. The focus of the present article falls therefore upon the concept of Lusophony as a place of reflection, of knowledge as well as recognition of oneself and the Other. This concept of Lusophony is instantiated by the notion of relation vis-à-vis globalisation and by the essence here portrayed by the linguistic ecological system of Portuguese.

A percepção da qualidade de voz por brasileiros bilíngues

Studies show that bilingual speakers can change their voices when they speak an L2 compared to their L1. It remains to be seen whether these differences are perceived by listeners. Thus, the present study addresses the perception of voice quality by Brazilian bilingual listeners in Brazilian Portuguese (PB) and English (IN) speech samples produced by Brazilian bilingual speakers. This objective was reached by means of a discrimination test of the PB and IN voices of the same speaker. The listeners judged whether the voices were the same or different, describing their characteristics if they were different. The results pointed to certain variability in the judgments, but also showed that listeners are able to identify differences in pitch and intensity between languages, as well as attribute personality and emotion characteristics to speech in both languages.

Escrita e leitura como práticas de resistência em Angola - a literatura nos anos que rondam a independência

Este artigo revisita um período da história angolana em que a nação começou a construir uma narrativa sobre si mesma, valendo-se da escrita e de leitura como práticas de resistência ao discurso português. Investiga ainda o resultado desse processo no pós-independência, quando o país vivenciou uma considerável euforia em torno do livro e as instituições à frente do processo prometiam uma pátria em que a educação para a literatura ocupava papel de destaque.

Global Accent in the Portuguese Speech of Heritage Returnees

The present study examined whether heritage speakers (HSs) of European Portuguese (EP) who were born or moved to a German-speaking country before the age of eight years were perceived as native speakers of EP. In particular, this study intended to determine whether a change of linguistic environment, length of residence in a migrant context, length of residence in the country of origin before migration and after remigration, and age at return could predict the degree of (non)native accent in the heritage language. Thirty native Portuguese speakers assessed the global accent of 20 Portuguese-German bilinguals, five Portuguese monolinguals and five highly proficient German speakers of Portuguese as a second language (L2). The group of HSs comprised 17 speakers who returned to Portugal. The results revealed that listeners perceived a strong global foreign accent in the speech of the L2 learners, while the monolingual Portuguese speakers were clearly perceived as being native speakers of EP. The HSs’ ratings were considerably closer to the monolingual average ratings, but they showed more variation, indicating that their accent may bear non-native traces. Further analyses showed that the age at which the HSs emigrated was the only significant predictor, while length of residence in the host country and in Portugal were less predictive.

Development of the speech test signal in Brazilian Portuguese for real-ear measurement

Objective: Recommended practice is to verify the gain and/or output of hearing aids with speech or speech-shaped signals. This study has the purpose of developing a speech test signal in Brazilian Portuguese that is electroacoustically similar to the international long-term average speech spectrum (ILTASS) for use in real ear verification systems. Design: A Brazilian Portuguese speech passage was recorded using standardized equipment and procedures for one female talker and compared to ISTS. The passage consisted of simple, declarative sentences making a total of 148 words. Study sample: The recordings of a Brazilian Portuguese passage were filtered to the ILTASS and compared to the International Speech Test Signal (ISTS). Aided recordings were made at three test levels, for three audiograms for the Brazilian Portuguese passage and the ISTS. Results: The unaided test signals were spectrally matched to within 0.5 dB. Aided evaluation revealed that the Brazilian Portuguese passage produced aided spectra that were within 1 dB on average, within about 2 dB per audiogram, and within about 3 dB per frequency for 95% of fittings. Conclusion: Results indicate that the Brazilian Portuguese passage developed in this study provides similar electroacoustic hearing-aid evaluations to those expected from the standard ISTS passage.

Tipologia de construções clivadas de um corpus de língua oral espontânea

Cleft constructions are studied under three broad approaches: generative, induced test or reading test and spontaneous speech. The objective of this article is to describe the typology of these constructions based on a spontaneous Brazilian Portuguese speech corpus. The results of this research show that clefting in spontaneous speech differs in some respects from introspective data analysis as well as from the previous descriptions of prosodic typologies induced by reading tests. The findings revealed besides prototypic cleft constructions, inverted, double copula, copula-less and interrogative constructions. It was also found that the pitch accent is not always in the non-presupposed part of these constructions.

Cochlear implant rehabilitation outcomes in Waardenburg syndrome children

ObjectivesThe purpose of this study was to review the outcomes of children with documented Waardenburg syndrome implanted in the ENT Department of Centro Hospitalar de Coimbra, concerning postoperative speech perception and production, in comparison to the rest of non-syndromic implanted children. MethodsA retrospective chart review was performed for children congenitally deaf who had undergone cochlear implantation with multichannel implants, diagnosed as having Waardenburg syndrome, between 1992 and 2011. Postoperative performance outcomes were assessed and confronted with results obtained by children with non-syndromic congenital deafness also implanted in our department. Open-set auditory perception skills were evaluated by using European Portuguese speech discrimination tests (vowels test, monosyllabic word test, number word test and words in sentence test). Meaningful auditory integration scales (MAIS) and categories of auditory performance (CAP) were also measured. Speech production was further assessed and included results on meaningful use of speech Scale (MUSS) and speech intelligibility rating (SIR). ResultsTo date, 6 implanted children were clinically identified as having WS type I, and one met the diagnosis of type II. All WS children received multichannel cochlear implants, with a mean age at implantation of 30.6±9.7months (ranging from 19 to 42months). Postoperative outcomes in WS children were similar to other nonsyndromic children. In addition, in number word and vowels discrimination test WS group showed slightly better performances, as well as in MUSS and MAIS assessment. ConclusionsOur study has shown that cochlear implantation should be considered a rehabilitative option for Waardenburg syndrome children with profound deafness, enabling the development and improvement of speech perception and production abilities in this group of patients, reinforcing their candidacy for this audio-oral rehabilitation method.

Polygonal Mesh Comparison Applied to the Study of European Portuguese Sounds

The purpose of the authors’ study was to evaluate the feasibility of using a mesh comparison tool in the study of European Portuguese speech sounds. A large 3D MRI database from several speakers, including various sounds and contexts has been acquired. Segmentation, visualization, and analysis of such a large database are complex, time-consuming tasks, preventing the use of manual segmentation techniques. A more efficient semi-automatic method was devised to accomplish that task. After tongue segmentation, meshes were created from the segmented volumes and polygonal mesh comparison was used to assess differences between different sounds, vocalic contexts, syllabic positions and speakers. This is the first study using such an approach to analyze and compare tongue shape. This comparison method provides a qualitative measure enabling further insight into the main differences between speech sounds.

On the development of verbal and nominal morphology in four lusophone creoles

AbstractThis article attempts to reconstruct the plausible evolution of inflectional and free-standing morphemes in three historically related lusophone creoles (Cape Verdean Creole, Guinea-Bissau Creole, Papiamentu), and compares their morphological properties to Angolar, believed to have followed an independent developmental path. I examine their synchronic morphological properties and seek insights into their origins by studying Black speech in Portuguese 15th- and 16th-century literature. This reconstruction seeks to address several questions. Within the historically related creoles, is it possible to identify a set of formally and functionally common morphemes? If so, what is their source? Are they likely to have been inherited through diffusion? Are the common morphemes in the three related creoles distinct from those of Angolar or do they overlap?

Bantu substratum interference in Mozambican Portuguese speech varieties

Bantu phonological constraints exert varying types of interference in the varieties of Portuguese spoken in Mozambique. While many of these processes – e. g. voicing dissimilation, consonant sequence simplification, and denasalization – are relatively well known among linguists specialized in the study of Mozambican Portuguese dialects, they have received only passing mention in the literature. Moreover, it is unclear from the available data how Bantu substratum influences characterize the Portuguese of L1 speakers and those who acquired the language at a very young age. The present work is based on the analysis of phonetic data from five native Portuguese speaking informants (three L1 speakers and two L2 speakers) from various locations in Mozambique, and whose language backgrounds represent Changana, Chope, GiTonga, Makhuwa, and Nyungwe. Evidence is presented showing the effects of local Bantu phonologies on informants’s spoken Portuguese varieties for the aforementioned phenomena, plus additional processes that correspond with the literature on Bantu linguistics, such as devoiced aspirated word-final consonants, ‘ whistled’ fricatives, and homorganic nasal epenthesis. We conclude that investigations of substratum phonetic and phonological interference are important for the long-term monitoring of a contact situation, and help to elucidate the principles governing how segments are realized in indigenizing languages.

Free tools and resources for Brazilian Portuguese speech recognition

Abstract An automatic speech recognition system has modules that depend on the language and, while there are many public resources for some languages (e.g., English and Japanese), the resources for Brazilian Portuguese (BP) are still limited. This work describes the development of resources and free tools for BP speech recognition, consisting of text and audio corpora, phonetic dictionary, grapheme-to-phone converter, language and acoustic models. All of them are publicly available and, together with a proposed application programming interface, have been used for the development of several new applications, including a speech module for the OpenOffice suite. Performance tests are presented, comparing the developed BP system with a commercial software. The paper also describes an application that uses synthesis and speech recognition together with a natural language processing module dedicated to statistical machine translation. This application allows the translation of spoken conversations from BP to English and vice versa. The resources make easier the adoption of BP speech technologies by other academic groups and industry.