“Accent” is the pattern of speech pronunciation by which one can identify a person’s linguistic, social or cultural background. It is an important source of inter-speaker variability and a particular problem for automated speech recognition. The aim of the study was to investigate a new computational approach to accent classification which did not require phonemic segmentation or the identification of phonemes as input, and which could therefore be used as a simple, effective accent classifier. Through a series of structured experiments this study investigated the effectiveness of Support Vector Machines (SVMs) for speech accent classification using time-based units rather than linguistically-informed ones, and compared it to the accuracy of other machine learning methods, as well as the ability of humans to classify speech according to accent. A corpus of read-speech was collected in two accents of English (Arabic and “Indian”) and used as the main datasource for the experiments. Mel-frequency cepstral coefficients were extracted from the speech samples and combined into larger units of 10 to 150ms duration, which then formed the input data for the various machine learning systems. Support Vector Machines were found to classify the samples with up to 97.5% accuracy with very high precision and recall, using samples of between 1 and 4 seconds of speech. This compared favourably with a human listener study where subjects were able to distinguish between the two accent groups with an average of 92.5% accuracy in approximately 8 seconds. Repeating the SVM experiments on a different corpus resulted in a best classification accuracy of 84.6%. Experiments using a decision tree learner and a rule-based classifier on the original corpus gave a best accuracy of 95% but results over the range of conditions were much more variable than those using the SVM. Rule extraction was performed in order to help explain the results and better inform the design of the system. The new approach was therefore shown to be effective for accent classification, and a plan for its role within various other larger speech-related contexts was developed.

“Accent” is the pattern of speech pronunciation by which one can identify a person’s linguistic, social or cultural background. It is an important source of inter-speaker variability and a particular problem for automated speech recognition. The aim of the study was to investigate a new computational approach to accent classification which did not require phonemic segmentation or the identification of phonemes as input, and which could therefore be used as a simple, effective accent classifier. Through a series of structured experiments this study investigated the effectiveness of Support Vector Machines (SVMs) for speech accent classification using time-based units rather than linguistically-informed ones, and compared it to the accuracy of other machine learning methods, as well as the ability of humans to classify speech according to accent. A corpus of read-speech was collected in two accents of English (Arabic and “Indian”) and used as the main datasource for the experiments. Mel-frequency cepstral coefficients were extracted from the speech samples and combined into larger units of 10 to 150ms duration, which then formed the input data for the various machine learning systems. Support Vector Machines were found to classify the samples with up to 97.5% accuracy with very high precision and recall, using samples of between 1 and 4 seconds of speech. This compared favourably with a human listener study where subjects were able to distinguish between the two accent groups with an average of 92.5% accuracy in approximately 8 seconds. Repeating the SVM experiments on a different corpus resulted in a best classification accuracy of 84.6%. Experiments using a decision tree learner and a rule-based classifier on the original corpus gave a best accuracy of 95% but results over the range of conditions were much more variable than those using the SVM. Rule extraction was performed in order to help explain the results and better inform the design of the system. The new approach was therefore shown to be effective for accent classification, and a plan for its role within various other larger speech-related contexts was developed.

Cette étude de l’accent secondaire en anglais britannique contemporain s’inscrit dans l’approche développée par Lionel Guierre et se compose d’un volet théorique et d’un volet empirique. Dans un premier temps, nous présentons cette approche et la comparons à d'autres théories contemporaines. Dans un second temps, après une revue de littérature, nous menons l'analyse d'un corpus de 5829 mots portant un accent secondaire. Cette étude confirme que la présence et la position de l'accent secondaire sont largement déterminées par des contraintres d'ordre rythmique. Elle révèle également des éléments nouveaux tels que le rôle joué par les préfixes sémantiquement opaques dans le placement de l'accent secondaire dans les mots dérivés et non-dérivés, le rôle de la fréquence relative entre dérivant et dérivé pour les dérivés accentués /021 (-) : ou encore l'existence d'une règle de détermination de la valeur de la voyelle accentuée dans la position initiale prétonique / This study of secondary accent in contemporary British English stands in the approach developed by Lionel Guierre and is composed of a theoretical section and an empirical section. First, we present this approach and compare it to other contemporary theories. Secondly, after a literature review, we conduct the analysis of a corpus of 5829 words carrying a secondary accent. This study confirms that the presence and position of secondary accent are largely determined by rhythmical contraints. It also reveals new elements such as the role of opaque prefixes in the placement of secondary accent in both derived an non-derived words, the role of the relative frequency of the base and its derivatives accented :021 (-)/or the existenza of a rule determining the value of the accented vowel in the initial pretonic position

Made available in DSpace on 2014-08-20T13:47:53Z (GMT). No. of bitstreams: 1 Luanda Alvariza Gomes Ney_Dissertacao.pdf: 1302951 bytes, checksum: 0084930b669bff09e5270f2e21767a98 (MD5) Previous issue date: 2012-04-13 / This research describes and analyses, in data extracted from spontaneous texts, the use of acute accent marks by young primary students. It is debated the relations established, in initial writing, between acute accent marks and children s knowledge about the language prosody. The proposal for attribution of acute accent marks by Bisol (1992, 1994) is the basis of this study. The analyzed corpus is comprised of 3161 words (not verbs) which receive acute accent marks, according to spelling norm, and 222 words where the words received acute accent marks when, according to the norm, it is not necessary. The resulting data were obtained by research made with 987 texts produced by children in two schools, a private and a public in the city of Pelotas/RS. The data were submitted to qualitative and quantitative analysis having into account two categories: correct answers and mistakes. The results demonstrate that: i) the majority of correct answers referring to acute accent marks occur in oxytone words and tonic monosyllables, mainly when the stressed vowels are the medium o and e ; ii) when stressing inadequately, children end up revealing coherent hypothesis related to phonetics and also to language prosody, because, in most of these cases, the acute accent marks, besides coinciding with the tonic syllable of the word, it is used for indicating open vowels; iii) the data of omission of acute accent marks represent 84% of the mistakes registered; iv) when the acute accent mark takes place in the changed syllable, it is verified a tendency in stressing as tonic the penultimate syllable, which characterizes the general rule of prosodic acute accent mark in Portuguese. The data confirm also that the progress in education constitutes an important factor for appropriation of rules of acute accent marks use by young learners, mainly, in private schools. Keywords: acute accent marks, prosodic accent marks, writing acquisition. / Esta pesquisa descreve e analisa, em dados extraídos de textos espontâneos, o uso da acentuação gráfica por crianças das séries iniciais. Discutem-se as relações estabelecidas, na escrita inicial, entre o acento gráfico e o conhecimento que as crianças possuem sobre a prosódia da língua. A proposta para atribuição do acento de Bisol (1992, 1994) serve de base para o estudo. O corpus analisado é constituído de 3161 palavras (não-verbos) que devem ser grafadas com acento, conforme a norma ortográfica da língua, e 222 palavras em que houve a grafia do acento quando este, segundo a norma, não é necessário. Os dados foram obtidos a partir do levantamento feito em 987 textos produzidos por crianças de duas escolas, uma pública e outra particular, da cidade de Pelotas/RS. Os dados foram submetidos à análise quali-quantitativa, considerando-se duas principais categorias: acertos e erros. Os resultados mostram que: i) o maior número de acertos referentes à grafia do acento ocorre em palavras oxítonas e monossílabos tônicos, principalmente quando as vogais acentuadas são as médias o e e ; ii) ao grafarem acentos indevidos, as crianças acabam por revelar hipóteses coerentes com a fonética e com a prosódia da língua, pois, na maioria desses casos, o acento, além de coincidir com a sílaba tônica da palavra, é utilizado para indicar vogais abertas; iii) os dados de omissão do acento representam 84% dos erros registrados; iv) quando ocorre grafia do acento na sílaba trocada, verifica-se uma tendência em marcar como tônica a penúltima sílaba, o que caracteriza a regra geral do acento prosódico do português. Os dados atestam ainda que o avanço na escolarização constitui um fator importante para a apropriação das regras de utilização do acento gráfico pelas crianças, sobretudo, na escola particular.

Abstract : In the contemporary world, English has become the international language in which most intercultural communication is conducted (Seidlhofer, 2011). The number of speakers using English as a foreign or second language (L2) outnumbers the number of native speakers of the language six times (Crystal, 2003) and, accordingly, English is used more commonly as a Lingua Franca in the international context than to communicate with its native speakers (Jenkins, 2007). Because of the global spread of English, there are more people who speak English with a non-native accent. Such accents often trigger a set of stereotypes insofar as it could have negative consequences for the speaker in terms of academic success and employability (Lippi-Green, 1994). From decades of research in social psychology and sociolinguistics, it has been established that accent plays a significant role in how native speakers perceive non-native speakers of English (Garrett, 2010). However, little has been done in exploring how non-native speakers of English view their own accent and judge acceptability of other non-native English accents. Therefore, this study addressed the need for further research focusing on non-native students’ attitudes towards Arabic- and Chinese-accented English in an English Intensive Program at an English-speaking university in Montreal. This study employed the verbal guise technique from the field of social psychology to elicit spontaneous reactions to L2 accented speech with two levels of first language (L1) influence. Unlike other verbal guise studies, this study did not include a native speaker sample for comparison. We believed our participants’ potential negative attitudes to L2-accented speech would be attenuated by the reference frame effect. To explain and clarify verbal guise test results, we also conducted semi-structured interviews. The results show that the respondents displayed a relatively low level of solidarity with Chinese and Arabic accents in English. Notwithstanding relatively low solidarity ratings given to the accented voices, they were still higher than the evaluations documented in the literature. In addition, high status ratings were found for non-native speakers despite the fact that they were recognized as such. This suggests that in addition to the positive effect triggered by the lack of inclusion of native speaker accents, there might be some contextual factors at play. We found that the participants were harsher in their ratings when the voices represented the more formal domain of higher education but more lenient when they referred to informal settings. From our data, it is also evident that non-segmentals play an important role in language attitudes. The female speakers who approximated their rhythm patterns and intonation contours to those of a native speaker received higher ratings on all the traits in our data. We believe that, if we want them to serve the needs of international students, these often ambivalent attitudes should be considered in the choice of a classroom pronunciation model, and in the development of pronunciation materials. / Résumé : Dans le monde contemporain, l’anglais est devenu la langue internationale utilisée pour la plupart des communications interculturelles (Seidlhofer, 2011). Le nombre de locuteurs utilisant l’anglais comme langue étrangère ou langue seconde est six fois plus important que le nombre de locuteurs natifs (Crystal, 2003). L’anglais est plus souvent utilisé comme Lingua Franca dans un contexte international que pour communiquer avec des locuteurs natifs (Jenkins, 2007). A cause de cette diffusion globale de l’anglais, la plupart des gens qui parlent anglais ont un accent non-natif. Ces accents déclenchent souvent une panoplie de stéréotypes qui peuvent avoir des conséquences négatives sur l’employabilité ou la réussite professionnelle des locuteurs (Lippi-Green, 1994). Des décennies de recherche en psychologie sociale et sociolinguistique ont reconnu que l’accent joue un rôle significatif sur la perception des locuteurs non-natifs par les locuteurs natifs (Garrett, 2010). Cependant, peu d’études se sont penchées sur la façon dont les locuteurs non-natifs perçoivent leur propre accent et jugent l’acceptabilité des autres accents non-natifs. Par conséquent, la présente étude se concentre sur les attitudes des étudiants internationaux du programme d’anglais intensif d’une université anglophone montréalaise envers les accents arabes et chinois en anglais. Cette étude utilise la technique de locuteur masqué employée dans le domaine de la psychologie sociale pour déclencher des réactions spontanées à un discours de langue seconde accentué par deux niveaux d’influence de la langue maternelle. Contrairement aux autres études utilisant la technique de locuteur masqué, cette étude n’inclus pas de discours de locuteurs natifs à des fins de comparaison. Nous pensons que les attitudes négatives potentielles pourraient être atténuées par l’effet de cadre de référence. Pour expliquer et clarifier les résultats de la technique de locuteur masqué, nous avons également mené des entrevues semi-dirigées. Les résultats montrent que les participants expriment un niveau de solidarité relativement faible envers les accents arabes et chinois en anglais. Pourtant, ce niveau de solidarité est tout de même plus haut que les études reportées dans la littérature. D’autre part, le statut social des locuteurs non-natifs a été évalué positivement malgré que les locuteurs aient été reconnus comme non-natifs. Ceci suggère qu’en plus d’un effet positif déclenché par l’exclusion des accents natifs, certains facteurs contextuels peuvent avoir influencé les résultats. Nous avons trouvé que les personnes interrogées notaient plus sévèrement les voix suggérant un niveau d’éducation élevé, mais étaient plus clémentes envers les voix faisant référence à un contexte informel. De l’analyse des données, il apparait que les non-segmentals jouent un rôle important dans les attitudes langagières. De plus, les locutrices se rapprochant du rythme et de l’intonation des natifs ont reçu des évaluations supérieures sur tous les traits. Nous pensons que ces attitudes souvent ambivalentes devraient être considérées dans le choix du modèle d’enseignement et dans le développement de nouveaux supports d’apprentissage de la prononciation pour les étudiants internationaux.

The status of English as the language of international communication is by now well-established. However, in the past 16 years, research has tried to emphasize the fact that the English spoken in international contact situations and between people with other first languages than English has different needs than the English spoken locally amongst native speakers, resulting in the emergence of English as a lingua franca (ELF) as a scholarly field. However, the impact of findings in ELF has so far only led to a moderate shift in English language teaching. Especially in expanding circle countries, where ELF should have the biggest impact, change is only gradually becoming palpable. Accent and pronunciation, as one of the biggest factors on both identity and mutual intelligibility (Jenkins 2000; 2007) are at the root of discussion. The scope of this study is therefore to examine accent choices and the extent to which native speaker ideology informs the preferences of ten speakers of ELF and 27 German natives with experience in international communication. Both ethnographical and sociolinguistic methods, as well as auditory analysis have been applied and conducted. The auditory analysis of six variables in the recorded speech production of the ten speakers suggests that there is no significant preference of one norm-giving variety over the other. Rather, speakers tend to mix-and-match General American- and Standard Southern British English-like features in their pronunciation. When reporting their accent ideals, the idea of a ‘neutral’ English accent is mentioned by four participants. Neutral accents seem to have been understood as ‘unmarked accents’. Expressed beliefs on their own English pronunciation show a comparatively high level of reflection on and confidence in their own production. Results from a rating task and a survey given to 27 German participants reveal attitudes that are more negatively stacked. While Germans reported openness towards NNS (non-native speaker) accents and showed awareness of the priority of intelligibility over accent choice in both their own and others’ pronunciation, they still largely reported NS accent preference. The ratings of the production from ten ELF speakers confirmed this and showed that ‘neutral’ is equated with native-like. In the light of these findings, issues are discussed that ultimately relate to the influence of NS Englishes, identity and the development of English as an international language.

Thesis (MScEng)--Stellenbosch University, 2012. / ENGLISH ABSTRACT: Several accents of English are spoken in South Africa. Automatic speech recognition (ASR) systems should therefore be able to process the di erent accents of South African English (SAE). In South Africa, however, system development is hampered by the limited availability of speech resources. In this thesis we consider di erent acoustic modelling approaches and system con gurations in order to determine which strategies take best advantage of a limited corpus of the ve accents of SAE for the purpose of ASR. Three acoustic modelling approaches are considered: (i) accent-speci c modelling, in which accents are modelled separately; (ii) accent-independent modelling, in which acoustic training data is pooled across accents; and (iii) multi-accent modelling, which allows selective data sharing between accents. For the latter approach, selective sharing is enabled by extending the decision-tree state clustering process normally used to construct tied-state hidden Markov models (HMMs) by allowing accent-based questions. In a rst set of experiments, we investigate phone and word recognition performance achieved by the three modelling approaches in a con guration where the accent of each test utterance is assumed to be known. Each utterance is therefore presented only to the matching model set. We show that, in terms of best recognition performance, the decision of whether to separate or to pool training data depends on the particular accents in question. Multi-accent acoustic modelling, however, allows this decision to be made automatically in a data-driven manner. When modelling the ve accents of SAE, multi-accent models yield a statistically signi cant improvement of 1.25% absolute in word recognition accuracy over accent-speci c and accentindependent models. In a second set of experiments, we consider the practical scenario where the accent of each test utterance is assumed to be unknown. Each utterance is presented simultaneously to a bank of recognisers, one for each accent, running in parallel. In this setup, accent identi cation is performed implicitly during the speech recognition process. A system employing multi-accent acoustic models in this parallel con guration is shown to achieve slightly improved performance relative to the con guration in which the accents are known. This demonstrates that accent identi cation errors made during the parallel recognition process do not a ect recognition performance. Furthermore, the parallel approach is also shown to outperform an accent-independent system obtained by pooling acoustic and language model training data. In a nal set of experiments, we consider the unsupervised reclassi cation of training set accent labels. Accent labels are assigned by human annotators based on a speaker's mother-tongue or ethnicity. These might not be optimal for modelling purposes. By classifying the accent of each utterance in the training set by using rst-pass acoustic models and then retraining the models, reclassi ed acoustic models are obtained. We show that the proposed relabelling procedure does not lead to any improvements and that training on the originally labelled data remains the best approach. / AFRIKAANSE OPSOMMING: Verskeie aksente van Engels word in Suid Afrika gepraat. Outomatiese spraakherkenningstelsels moet dus in staat wees om verskillende aksente van Suid Afrikaanse Engels (SAE) te kan hanteer. In Suid Afrika word die ontwikkeling van spraakherkenningstegnologie egter deur die beperkte beskikbaarheid van geannoteerde spraakdata belemmer. In hierdie tesis ondersoek ons verskillende akoestiese modelleringstegnieke en stelselkon gurasies ten einde te bepaal watter strategie e die beste gebruik maak van 'n databasis van die vyf aksente van SAE. Drie akoestiese modelleringstegnieke word ondersoek: (i) aksent-spesi eke modellering, waarin elke aksent apart gemodelleer word; (ii) aksent-onafhanklike modellering, waarin die akoestiese afrigdata van verskillende aksente saamgegooi word; en (iii) multi-aksent modellering, waarin data selektief tussen aksente gedeel word. Vir laasgenoemde word selektiewe deling moontlik gemaak deur die besluitnemingsboom-toestandbondeling-algoritme, wat gebruik word in die afrig van gebinde-toestand verskuilde Markov-modelle, uit te brei deur aksent-gebaseerde vrae toe te laat. In 'n eerste stel eksperimente word die foon- en woordherkenningsakkuraathede van die drie modelleringstegnieke vergelyk in 'n kon gurasie waarin daar aanvaar word dat die aksent van elke toetsspraakdeel bekend is. In hierdie kon gurasie word elke spraakdeel slegs gebied aan die modelstel wat ooreenstem met die aksent van die spraakdeel. In terme van herkenningsakkuraathede, wys ons dat die keuse tussen aksent-spesi eke en aksent-onafhanklike modellering afhanklik is van die spesi eke aksente wat ondersoek word. Multi-aksent akoestiese modellering stel ons egter in staat om hierdie besluit outomaties op 'n data-gedrewe wyse te neem. Vir die modellering van die vyf aksente van SAE lewer multi-aksent modelle 'n statisties beduidende verbetering van 1.25% absoluut in woordherkenningsakkuraatheid op in vergelyking met aksent-spesi eke en aksent-onafhanklike modelle. In 'n tweede stel eksperimente word die praktiese scenario ondersoek waar daar aanvaar word dat die aksent van elke toetsspraakdeel onbekend is. Elke spraakdeel word gelyktydig gebied aan 'n stel herkenners, een vir elke aksent, wat in parallel hardloop. In hierdie opstelling word aksentidenti kasie implisiet uitgevoer. Ons vind dat 'n stelsel wat multi-aksent akoestiese modelle in parallel inspan, e ense verbeterde werkverrigting toon in vergelyking met die opstelling waar die aksent bekend is. Dit dui daarop dat aksentidenti seringsfoute wat gemaak word gedurende herkenning, nie werkverrigting be nvloed nie. Verder wys ons dat die parallelle benadering ook beter werkverrigting toon as 'n aksent-onafhanklike stelsel wat verkry word deur akoestiese en taalmodelleringsafrigdata saam te gooi. In 'n nale stel eksperimente ondersoek ons die ongekontroleerde herklassi kasie van aksenttoekennings van die spraakdele in ons afrigstel. Aksente word gemerk deur menslike transkribeerders op grond van 'n spreker se moedertaal en ras. Hierdie toekennings is nie noodwendig optimaal vir modelleringsdoeleindes nie. Deur die aksent van elke spraakdeel in die afrigstel te klassi seer deur van aanvanklike akoestiese modelle gebruik te maak en dan weer modelle af te rig, word hergeklassi seerde akoestiese modelle verkry. Ons wys dat die voorgestelde herklassi seringsalgoritme nie tot enige verbeterings lei nie en dat dit die beste is om modelle op die oorspronklike data af te rig.

This literature review sets out to revisit the studies exploring impact of rater characteristics on language oral assessments. Three categories of raters' backgrounds: occupation, accent familiarity, and native language are identified and will be addressed respectively in the following sections. The results showed that no consensus regarding raters' occupational background, linguistic background and native-speaker status on examiners' rating has been found so far. However, this review will highlight the current testing situations, bring up limitations from previous studies, provide implications for both teachers and raters, and hopefully shed light on future research. / text

The major aim of the thesis is to study the development of French spelling since the 16th century with main focus on the diacritical mark accent circonflexe. The individual centuries are analysed step by step from the linguistic perspective with the historical context addition. The most significant pieces dealing with the French spelling from the period of a formation of spelling rules to the current period of spelling reforms and their consequences in 2016 are analysed. This thesis focuses on the development of the mark accent circonflexe in detail. The main intention of the thesis is to describe the introduction of this mark to the French spelling, its struggle for survival, an extension and gaining its importance till its official status when it becomes the French spelling icon threatened by the reform of spelling in 1990 proposing the restriction of this mark. The research part in the form of a questionnaire is added to the master thesis. The opinion of native francophone people on the reform of spelling, especially on accent circonflexe is under examination.

The study focuses on durational variation of segments in read speech of Czech and British speakers of English. The variation in segmental duration was examined in the speech of three Czech and three British speakers with respect to individual vowel and consonant categories. Furthermore, stress and intonation phrase boundary were explored as possible factors influencing segmental duration. The theoretical part of the study focuses on two main topics. The first is foreign accent with regard to accentedness, intelligibility and comprehensibility as well as in connection to its implications and factors which might influence its degree. The second part contains theoretical background about the time domain of speech concerning segmental duration in Czech and English, the area of rhythm, the importance of durational cues in perception and a short survey of factors reported to affect segmental duration. Czech accent in English and the importance of duration in foreign language are also covered. In the empirical part of the study, Czech speakers were found to miss the appropriate proportions between individual vowel categories and realize their long vowels as slightly too short. They lacked an adequate degree of durational contrast in stressed and unstressed vowels when compared to the British speakers. With regard...