Multilingual Speaker Research Articles

Language Dependency of /s/ Production: Native Dutch Versus Non-Native English.

With forensic recordings being collected in multiple languages increasingly often, this study investigates the language dependency of the voiceless alveolar fricative /s/ in speakers of native (L1) Dutch and non-native (L2) English. Due to phonetic similarity between the languages, Dutch learners of English may exhibit language-independent /s/ acoustics, making it an interesting feature for multilingual forensic speaker comparisons (FSCs). However, the findings show that out of the four spectral moments, center of gravity, standard deviation (SD), skewness, and kurtosis, only SD remained stable across the languages; the other measurements were language-dependent. The results were largely independent of the /s/ tokens' contexts, although an interaction between language and context was found for skewness and kurtosis: With a labial right phonetic neighbor, language dependency was largely reduced. The findings have implications for FSCs: as /s/ is language-dependent in speakers of L1 Dutch and L2 English, it shows limited potential for cross-linguistic speaker comparisons in forensic casework.

Recruiters’ beliefs regarding the employability of native and non-native English-speaking tertiary-level teachers: The case of public universities in Thailand

This mixed-method study explores the criteria that recruiters in Thai public universities consider important when hiring English language teachers and the importance they attach to the native speaker criterion. The data were collected through an online questionnaire with ninety-four recruiters and semi-structured interviews with ten recruiters. The questionnaire data revealed that the participants considered ten criteria important, including English language proficiency, performance in an interview, teaching demonstration, educational background, teaching qualification, area of academic expertise, application documents, pronunciation, teaching experience, and visa status. Moreover, the participants viewed the native speaker criterion as somewhat important. The interview data yielded rich data about the thinking behind the use of the criteria. Particularly, it showed different positions towards the native speaker criterion, ranging from the strong preference towards native English-speaking candidates to the rejection of the native/non-native dichotomy which favors the competent bi-/multilingual speaker of English as a target and pedagogical model.

Real-time multilingual speech recognition and speaker diarization system based on Whisper segmentation.

This research presents the development of a cutting-edge real-time multilingual speech recognition and speaker diarization system that leverages OpenAI's Whisper model. The system specifically addresses the challenges of automatic speech recognition (ASR) and speaker diarization (SD) in dynamic, multispeaker environments, with a focus on accurately processing Mandarin speech with Taiwanese accents and managing frequent speaker switches. Traditional speech recognition systems often fall short in such complex multilingual and multispeaker contexts, particularly in SD. This study, therefore, integrates advanced speech recognition with speaker diarization techniques optimized for real-time applications. These optimizations include handling model outputs efficiently and incorporating speaker embedding technology. The system was evaluated using data from Taiwanese talk shows and political commentary programs, featuring 46 diverse speakers. The results showed a promising word diarization error rate (WDER) of 2.68% in two-speaker scenarios and 11.65% in three-speaker scenarios, with an overall WDER of 6.96%. This performance is comparable to that of non-real-time baseline models, highlighting the system's ability to adapt to various complex conversational dynamics, a significant advancement in the field of real-time multilingual speech processing.

THE NO TRAIN NO GAIN SYSTEM FOR O-COCOSDA AND VLSP 2022 - A-MSV SHARED TASK: ASIAN MULTILINGUAL SPEAKER VERIFICATION

This paper proposes a semi-supervised multilingual speaker verification (MSV) system submitted for the 2 tasks, MSV for the Asian language inside the training set (T01) and outside the training set (T02) in O-COCOSDA and VLSP challenge 2022.To solve the problem, our strategy is training a baseline acoustic model with given labeled data (MSV CommonVoice) andfine-tuning the trained acoustic model with both given labeled data and given unlabeled data (MSV Youtube). To achieve the fine-tuning step, the unlabeled data is converted to labeled data by pseudo labeling technique using the clustering method with the embedding vectors extracted from the trained acoustic model. Besides, we also apply test-time augmentation, back-end scoring, and score normalization with the AS-Norm technique to improve the result. When evaluated on the VLSP 2022 challenge's given test set, our best system with baseline ECAPA-TDNN achieves an equal error rate (EER) of 2.296% in T01 and 3.3296% in T02, which ranks second rank in both two tasks.

Multilingual Speaker Recognition using Mel-frequency Cepstral Coefficients and Gaussian Mixture Model

Introduction: People can recognize a speaker with the help of their voice via mobile or digital devices. Method: To obtain this congenital human being ability, authentication techniques based on speaker biometrics like automated speaker recognition (ASR) have been proposed. An ASR identifies speakers by speech signals analysis and salient feature extraction from their voices. Result: This will become an important part of recent research in the voice biometrics field. This paper proposes multilingual speaker recognition with the help of MFCC as feature extraction and GMM as classification techniques using various available datasets such as TIMIT, librespeech, etc. Conclusion: The results achieved from the given datasets enhance the recognition rate of 70.98% with MFCC.

Improving the Performance of Low-resourced Speaker Identification with Data Preprocessing

Automatic speaker identification is done to tackle daily security problems. Speech data collection is an essential but very challenging task for under-resourced languages like Burmese. The speech quality is crucial to accurately recognize the speaker’s identity. This work attempted to find the optimal speech quality appropriate for Burmese tone to enhance identification compared with other more richy resourced languages on Mel-frequency cepstral coefficients (MFCCs). A Burmese speech dataset was created as part of our work because no appropriate dataset available for use. In order to achieve better performance, we preprocessed the foremost recording quality proper for not only Burmese tone but also for nine other Asian languages to achieve multilingual speaker identification. The performance of the preprocessed data was evaluated by comparing with the original data, using a time delay neural network (TDNN) together with a subsampling technique that can reduce time complexity in model training. The experiments were investigated and analyzed on speech datasets of ten Asian languages to reveal the effectiveness of the data preprocessing. The dataset outperformed the original dataset with improvements in terms of equal error rate (EER). The evaluation pointed out that the performance of the system with the preprocessed dataset improved that of the original dataset.

“I made many discoveries for myself”: The development of a teacher candidate's pedagogical knowledge of translanguaging

While research has highlighted the importance of translanguaging for emergent multilingual learners (EMLs), very few studies have focused on teacher candidates (TCs)’ knowledge of translanguaging and how they implement the pedagogy. Using ethnographic methods and multimodal conversation analysis, this study examined how an English as an Additional Language (EAL) TC's knowledge of translanguaging evolved and how she bridged theory to practice in a field-based setting. The participant was a multilingual speaker who did not speak her EMLs' home languages. Data sources included: 1) the participant's course assessments, 2) a reflection, and 3) transcripts of a video classroom observation. The findings revealed her strong stance and ability to use a variety of strategies to create a translanguaging space, as a result of her training. We recommend teachers to enrich their pedagogical repertoire by understanding the affordances of a diverse range of multilingual, multimodal, and multi-sensory resources to effectively facilitate the teaching and learning processes. We also advise teacher educators to be intentional in course design, offering opportunities for TCs to reflect and connect their second language acquisition knowledge with pedagogical translanguaging in order to discover their own ways to create a translanguaging space for EMLs.

The Impact of Codeswitching in a Multilingual Speaker

The aim for this paper is to present the results of a research conducted to identify the impact of codeswitching in a multilingual speaker. The research was conducted via electronic since the participant and the researcher are not in the same country. The results were obtained by an interview where the participant share her experience by being a multilingual speaker in Germany. The results show the effect and how the participant feel regarding the phenomenon of codeswitching.

Bringing Race to the Classroom: How a Multilingual Speaker Performs Infra Politics to Confront Raciolinguistic Ideologies

The notion of (anti) racism in applied linguistics in general and in language education in particular has gained considerable attention by scholars in the fields. Contesting the dominance of monolingualism in language education, applied linguists and language education scholars have become eager to resuscitate this notion, often implicitly averring that racism has long been insidiously penetrating in the field and surreptitiously operating under the so-called raciolinguistic ideologies. It is these ideologies that are alleged to perpetuate, and even to further the hegemony of White supremacy and empire, eventually giving rise to racial inequalities and racial hierarchies in language education. The antiracism movement, it has been asserted, needs to be enacted. This article will argue that the fervent pronouncements of raciolinguistic ideologies need to be taken seriously, so as to promote linguistic justice and linguistic equality in language education. It will first discuss the claims of raciolinguistic ideologies, and then provide examples (from a classroom practice) of how the so-called “racialized subjects” enact their quiescent capacity as social and political being in subverting identities in the perceived dominant language (i.e. English) as a way of doing infra politics –an instance of grassroots politics. In so doing, the article argues that the racialized subjects are not submissive language users, but are actively engaged themselves in resisting raciolinguistic ideologies

A comparative analysis of neural network function: resilient back propagation algorithm (BPA) and radial basis functions (RBF) in multilingual environment

The most convenient speech processing tool is Artificial Neural Networks (ANNs). The effectiveness has been tested with various real-time applications. The classifier using artificial neural networks identifies utterances based on features extracted from the speech signal. The proposed approach to multilingual speaker identification consists of two parts, such as a training part and a testing part. In the training part, the classifier is trained using speech feature vectors. The spoken language contains complete information, such as details about the content of the message and details about the speaker of that message. In the present work, the speech signal databases of different speakers in a multilingual environment were recorded in three Indian languages, i.e., Hindi, Marathi, and Rajasthani. The cepstral characteristics of the speech signal were extracted: Mel-Frequency Cepstral Coefficients (MFCC) and Gammatone Frequency Cepstral Coefficients (GFCC). The system is designed for speaker recognition through multilingual speech signals using MFCC, GFCC, and combined functions as acoustic characteristics. Training and testing were performed using the Neural Network (NN) function, robust Backpropagation Algorithm (BPA), and Radial Basis Functions (RBF), and the results were compared. The accuracy of the speaker identification system is 94.89% using BPA and 96.62% using the RBF neural network.

Multi-lingual Speaker Recognition based on Asymmetric Convolution and Central Loss Function

In order to solve the problem of poor speaker recognition performance on multi-language corpus on convolutional neural networks and large amount of calculations on multiple parameters, this paper draws on asymmetric convolution and center loss (Center Loss) functions to improve and optimize the ResNet model. And perform speaker recognition tasks in the Chinese-Uyghur corpus. The results show that, compared with the original model, the improved model has higher accuracy, fewer parameters, and reduced computational complexity.

Learning Optimum Number of Bases for Indian Languages in Non-negative Matrix Factorization based Multilingual Speech Separation

Non-negative matrix factorization-based audio source separation separating a target source has shown significant performance improvement when the spectral bases attained after factorization exhibits latent structures in the mixed audio signal comprising multiple speaker sources. If all the sources are known, the spectral bases may be inferred on priority by using a training process on the database of isolated sources. The number of bases inferred for a source should not include bases matching spectral patterns of the interfering sources in the audio mixture; otherwise, the estimated target source after separation will be incorporated with undesirable spectral patterns. It is difficult to distinguish and separate similar audio sources in an overlapped speech, leading to a complex speech processing task. Therefore, this research attempts to learn an optimum number of bases for Indian languages leading to successful separation of target source in multi-lingual multiple speaker speech mixtures using non-negative matrix factorization. The languages used for utterances are Hindi, Marathi, Gujarati, and Bengali. The speaker combinations used are female-female, male-male, and female-male. The optimum number of bases which was determined by evaluating improvement in the separation performance was found to be 40 for all the languages considered.

A Novel Approach to Increase the Efficiency of a Multi-lingual Real-time Speaker Identification System

Nowadays, the real-time speaker recognition system is very popular due to its cost-effective nature. However, it is a very challenging one to produce a more efficient speaker identification system. In our work, we work on a multi-lingual real-time speaker identification system. We work in a novel way to enhance the efficiency of the said system. We take some real speech signals and use different speech enhancement methods and our proposed voice activity method (VAD) to enhance the efficiency of said system. By doing so, we increase the accuracy of the said system relatively by 2% as compared to existing methods.

Who’s Afraid of Multilingualism? Language and Intersectionality

This contribution to the special issue addresses questions of social justice in the field of applied linguistics through the optics of multilingualism and intersectionality and from the standpoint of language teachers and learners. In monolingualist ideology, the multilingual speaker is transformed from the normal condition to the linguistic other, especially those without a clearly identifiable L1 or whose L1 does not conform to their presupposed linguistic identity. Linguistic theory contributes to this result by assigning an almost mystic quality to the “mother tongue” and its “native speakers,” when these questions can be and are quite fluid in our multilingual world. This article suggests the need for paradigm shifts in three areas. First, the too-frequent focus on how to best promote learning in the target language needs to make way for how to best promote multilingual proficiency up to advanced knowledge of several languages. Second, it is time to demystify the “mother tongue” or “native” or “first” language. Finally, the field needs to conceptualize “non-native speaking” teachers for their essential qualities—their multilingual proficiency—to form a better valuation of their actual language teaching skillset.

The virtue of imperfection. Gjorgji Pulevski’s Macedonian–Albanian–Turkish dictionary (1875) as a window into historical multilingualism in the Ottoman Balkans

AbstractEven though the Balkans constitute one of the most prominent examples of linguistic areas, little is known about the actual processes and mechanisms contributing to the shaping of this area. Most of the assumptions are based on macro-level analyses and describe the linguistic changes observed in terms of generalising tendencies such as increase in analytism or simplification of structures. In order to approach the processes underlying contact-driven change and area formation, however, the focus needs to be shifted to the micro-level, i. e. the individuals and their communicative practices. Among the rare sources allowing to assume this actor-centred perspective is Gjorgji Pulevski’s trilingual dictionary of Macedonian, Albanian and Turkish from 1875, which on the orthographic and morphological level allows for insight into a multilingual speaker’s perceptions the languages s/he is exposed to and makes use of in her/his every day communicative practice. The present paper discusses the structural parallels between Macedonian and Turkish observed in the dictionary. It illustrates in how far these parallels may contribute to our understanding of the specific kind of individual multilingualism that provided the basis for the morphosyntactic developments observed for the Balkan linguistic area, and may also help to shed light on the more general nature of these developments. It is suggested that these processes evince an increase in morphological transparency, i. e. morphem-to-function mapping, as the most salient and probably most effective outcome of largely imperfect multilingualism.

Campur Kode Ceramah Ustadzah Mumpuni Handayani: Analisis Sosiolinguistik

This study aims to determine the use of code mixing in the video study of Ustazah Mumpuni Handayayekti by using the analysis of sociolinguistic theories. The results of this study are using mixed code in lecturing Ustadzah Mumpuni Handayayekti into the use of: (1) mixed code of Javanese word forms, (2) mixed code of Javanese clause form codes, (3) mixed code of Balinese word forms, and (4) mixed code phrase of Javanese language. In addition, mixing the code out on the use (1) mixed form codes for Arabic clauses, (2) mixed form codes for Arabic words, (3) mixed form codes for Arabic phrases, and (4) mixed form codes for Indonesian words. Mixing the code in and mixing the code out used by Ustadzah Mumpuni Handayayekti in his lecture, shows that he is a multilingual speaker and can use variations of language well in the right context.

“It is a constant struggle of trying to ‘stick to my roots’”: a narrative analysis of an immigrant’s identities-in-making

ABSTRACT Informed by the theory of identity in practice, this study examined an immigrant student’s (Yulia’s) language learning experience. A narrative analysis of Yulia’s personal stories reveals that her perceived identities as an intelligent, different, and special multilingual speaker have isolated her from her communities. She desperately tried to fit in with various groups in multiple communities of practice. Unfortunately, although she felt a sense of belonging, she became a modified version of herself. Shuttling among languages, cultures, and groups, Yulia found that her true self was invisible in any culture in which she found herself. Moreover, positioned as her parents’ secretary and translator, Yulia believed that her lived experiences have shaped her as a resilient, mature, and independent immigrant who shouldered many familial responsibilities at a young age. Her reversed role as a parent deprived her of time and joy to be a child. Yulia’s multilayered and complicated identities interacted with her actions and perspectives on herself and experiences as an immigrant. Findings indicate that this immigrant’s self-formation in action at times matches and mismatches her self-conceptions based on her learning trajectories and unique situations. Thus, the researcher argues that linguistic competence does not necessarily facilitate community membership establishment.

Language exposure in early bilingual and trilingual acquisition

ABSTRACTThe present study contributes to the field of bilingual and trilingual language acquisition by analysing four multilingual families. In each family, at least one parent is bilingual. Following research questions will be addressed: (A) How relevant are the (amount of) input and the (kind of) parental language strategies for language maintenance? (B) Which language choice can we observe in bilingual parents? (C) Is there a uniform definition for multilingual acquisition of immigrant languages? (D) Which language-external factors are possibly relevant for language transmission? The results reveal that exposure in some languages is lower than 20%, a level regarded as possibly crucial for becoming a productive multilingual speaker (Quay [2008]. Dinner conversations with a trilingual two-year-old: Language socialisation in a multilingual context. First Language, 28(1), 5–33). Some bilingual parents use the minority, others the majority language. Independently of the acquisition context, we claim it is unnecessary to label second- and third-generation children heritage language speakers. Taking the family language policy approach, we believe language practices in families with at least one bilingual parent are largely understudied and should be analysed more carefully.

중국어 학습동기 변화에 관한 일기연구: 행위이론 관점에서

Despite the extensive research on second language learning motivation, very little research has been conducted with Chinese learning motivation and the focus of the research has been limited mainly to identifying types of motivation. Considering the popularity of Chinese as a second language and the lack of research on Chinese learning motivation, the study aims to illustrate the process of motivational changes in learning Chinese from the sociocultual perspective. The data were collected from the researcher’s own autobiographic diary written while she was learning Chinese. Using the framework of an Activity Theory, the data were analyzed to find factors affecting the participant’s motivatonal changes. The findings indicate that the participant’s motivation was affected by dynamic interaction of mediational tools, relationships within the community, and different roles of the learner. The findings imply that the level of motivation is affected by how the learner resolves the tensions among those factors to move forward to the goal of becoming a multilingual speaker. The study may provide a methodological insight into L2 motivation research by analyzing and describing a longitudinal Chinese learning experience from a first-person point of view.

A multi-direcionalidade da transferência da duração vocálica do português como L3 para a L1 (espanhol) e a L2 (inglês): um estudo longitudinal

Este trabalho, à luz da Teoria dos Sistemas Dinâmicos, Adaptativos Complexos (BECKNER et al., 2009; DE BOT et al., 2013; SILVA, 2014), visa a investigar a premissa de que uma mudança em qualquer parte do sistema linguístico pode gerar alteração nas suas outras partes componentes. A partir desta premissa, conduziu-se um estudo de cunho longitudinal. O estudo contou com um aprendiz trilíngue, residente no Brasil, falante do espanhol (variedade mexicana) como L1, inglês como L2 e português como L3. Tal estudo contou com instrução formal, de base comunicativa e articulatória, sobre sons vocálicos do Português Brasileiro (L3). Através da instrução formal, visou-se a causar uma modificação acelerada no sistema fonético-fonológico de L3 do aprendiz, para se verificar se tal modificação ocasionaria efeitos sobre a sua L1 (Espanhol) e a sua L2 (Inglês). Verificou-se, para esse fim, a duração das vogais tônicas nas três línguas. A instrução deu-se ao longo de 4 meses, com uma aula de 90 minutos por semana. As coletas contaram com listas de palavras nas três línguas referidas, e ocorreram anteriormente ao período de instrução, durante (a cada 4 semanas) e ao término da instrução formal, de modo a totalizar 5 coletas com o participante. Os achados deste estudo longitudinal evidenciaram que as alterações nas durações das vogais do português (L3), aceleradas pela instrução fornecida, causaram alterações nas durações vocálicas (absoluta e relativa) das demais línguas, devido à interconexão dos sistemas do falante multilíngue.