published_or_final_version / abstract / toc / Linguistics / Master / Master of Arts

published_or_final_version / abstract / toc / Linguistics / Master / Master of Arts

published_or_final_version / abstract / toc / Linguistics / Master / Master of Arts

published_or_final_version / abstract / toc / Linguistics / Master / Master of Philosophy

While text Information Retrieval applications often focus on extracting semantic features to identify the topic of a document, and Music Information Research tends to deal with melodic, timbral or meta-tagged data of songs, useful information can be gained from surface-level features of musical texts as well. This is especially true for texts such as song lyrics and poetry, in which the sound and structure of the words is important. These types of lyrical verse usually contain regular and repetitive patterns, like the rhymes in rap lyrics or the meter in metrical poetry. The existence of such patterns is not always categorical, as there may be a degree to which they appear or apply in any sample of text. For example, rhymes in hip hop are often imperfect and vary in the degree to which their constituent parts differ. Although a definitive decision as to the existence of any such feature cannot always be made, large corpora of known examples can be used to train probabilistic models enumerating the likelihood of their appearance. In this thesis, we apply likelihood-based methods to identify and characterize patterns in lyrical verse. We use a probabilistic model of mishearing in music to resolve misheard lyric search queries. We then apply a probabilistic model of rhyme to detect imperfect and internal rhymes in rap lyrics and quantitatively characterize rappers' styles in their use. Finally, we compute likelihoods of prosodic stress in words to perform automated scansion of poetry and compare poets' usage of and adherence to meter. In these applications, we find that likelihood-based methods outperform simpler, rule-based models at finding and quantifying lyrical features in text.

This study examines the production and perception of laryngeal constriction in the early vocalizations of Bai and English infants. The first part of the study documents the development of laryngeal voice quality features in the non-syllabic and syllabic utterances of Bai and English infants. The second part of the study focuses on the perception of laryngeal constriction in infant vocalizations by adult Bai and English listeners. The study is grounded in Esling’s (2005) model of the vocal tract, which characterizes the laryngeal vocal tract as a separate articulator, distinct from the oral vocal tract. The study of Bai and English infants’ production identifies universal and language-specific patterns in infants’ development of laryngeal constriction. In the first months of life, most sounds produced by Bai and English infants are constricted. As the year progresses, all infants explore degrees of constriction in dynamic utterances that feature alternations between constricted and unconstricted laryngeal voice quality settings. As well, throughout the year, infants produce an increasing proportion of unconstricted vocalizations. By the end of the first year, when infants have developed increasing control of the laryngeal and oral vocal tracts, they produce syllabic utterances that begin to reflect the use of laryngeal voice quality features in their ambient language. English syllabic utterances are mostly unconstricted, mirroring the prevalence of unconstricted settings in the target language. By contrast, Bai syllabic utterances are mostly constricted or dynamic, reflecting the use of laryngeal voice quality in Bai, a register tone language that employs laryngeal voice quality features distinctively at the syllabic level. The second part of the study highlights universal and language-particular patterns in Bai and English adults’ perception of laryngeal voice quality in infants’ utterances. In evaluating the importance of a range of infant sounds in learning the target language (Bai or English), adults from both language groups assign lower ratings to infant utterances that occur earlier in development, such as constricted non-syllabic utterances, and higher ratings to sounds that occur later, such as syllabic utterances with rapidly articulated syllables. Bai and English adults’ perceptions also reflect some language-specific patterns that correspond to language-particular characteristics identified in infants’ use of laryngeal voice quality in syllabic and non-syllabic utterances. These correspondences suggest that adults are attuned to laryngeal voice quality in infants, and that, in turn, infants become attuned to the use of laryngeal voice quality features in their ambient language early in development. The production study demonstrates the fruitfulness of Esling’s (2005) model of the vocal tract in revealing previously undocumented patterns in the development of laryngeal constriction in the first year of life and in highlighting the importance of emergent laryngeal control as a stimulator of phonetic development. The perception study shows that adults whose native languages differ markedly in their use of laryngeal constriction can systematically evaluate laryngeal voice quality features in the full range of non-distress vocalizations produced by infants in the first year of life.

Speaker verification is the process of verifying or disputing the claimed identity of a speaker based on a recorded sample of their speech. Automatic speaker verification technology can be applied to a variety of person authentication and identification applications including forensics, surveillance, national security measures for combating terrorism, credit card and transaction verification, automation and indexing of speakers in audio data, voice based signatures, and over-the-phone security access. The ubiquitous nature of modern telephony systems allows for the easy acquisition and delivery of speech signals for processing by an automated speaker recognition system. Traditionally, approaches to automatic speaker verification have involved holistic modelling of low-level acoustic-based features in order to characterise physiological aspects of a speaker such as the length and shape of the vocal tract. Although the use of these low-level features has proved highly successful, there are numerous other sources of speaker specific information in the speech signal that have largely been ignored. In spontaneous and conversational speech, perceptually higher levels of in- formation such as the linguistic content, pronunciation idiosyncrasies, idiolectal word usage, speaking rates and prosody, can also provide useful cues as to identify of a speaker. The main aim of this work is to explore the incorporation of higher levels of information into the verification process. Specifically, linguistic constructs such as words, syllables and phones are examined for their usefulness as features for text-independent speaker verification. Two main approaches to incorporating these linguistic features are explored. Firstly, the direct modelling of linguistic feature sequences is examined. Stochastic language models are used to model word and phonetic sequences obtained from automatically obtained transcripts. Experimentation indicates that significant speaker characterising information is indeed contained in both word and phone-level transcripts. It is shown, however, that model estimation issues arise when limited speech is available for training. This speaker model estimation problem is addressed by employing an adaptive model training strategy that significantly improves the performance and extended the usefulness of both lexical and phonetic techniques to short training length situations. An alternate approach to incorporating linguistic information is also examined. Rather than modelling the high-level features independently of acoustic information, linguistic information is instead used to constrain and aid acoustic- based speaker verification techniques. It is hypothesised that a ext-constrained" approach provides direct benefits by facilitating more detailed modelling, as well as providing useful insight into which articulatory events provide the most useful speaker-characterising information. A novel framework for text-constrained speaker verification is developed. This technique is presented as a generalised framework capable of using di®erent feature sets and modelling paradigms, and is based upon the use of a newly defined pseudo-syllabic segmentation unit. A detailed exploration of the speaker characterising power of both broad phonetic and syllabic events is performed and used to optimise the system configuration. An evaluation of the proposed text- constrained framework using cepstral features demonstrates the benefits of such an approach over holistic approaches, particularly in extended training length scenarios. Finally, a complete evaluation of the developed techniques on the NIST2005 speaker recognition evaluation database is presented. The benefit of including high-level linguistic information is demonstrated when a fusion of both high- and low-level techniques is performed.

Ce travail s’inscrit dans le cadre global de la didactique du FLE/FLS puis dans celui, plus précis, de la prononciation du français parlé et porte sur les difficultés des apprenants hungarophones en production orale. Si le matériel phonique des deux langues présente certaines similitudes, le point nodal des interférences réside surtout dans les oppositions vocaliques (durée phonologique en hongrois vs durée phonétique en français), la prononciation de certaines consonnes et la prosodie (accent tonique d’intensité hongrois vs accent rythmique duratif français). Diverses méthodes expérimentales usitées à l’Institut de Phonétique de Strasbourg seront exploitées : spectrographie pour le timbre et la durée des voyelles ainsi que pour les phénomènes prosodiques, laryngographie pour les assimilations de sonorité et contacts consonantiques. Les analyses se situeront donc au niveau segmental, comme au niveau suprasegmental. Outre de se poser en tant que contribution à l’étude comparée et contrastive des langues, cette recherche privilégiera les visées didactique et pédagogique et apportera en dernier lieu explications et astuces de remédiation. Nous espérons que cette thèse sera utile à nos collègues français exerçant en Hongrie, mais aussi à nos collègues hongrois n’ayant pas forcément conscience de certaines fautes typiques et redondantes, car partant du même substrat que leurs apprenants. / This research fits into the overall framework of FLE/FLS teaching methods and more precisely, of spoken French pronunciation. It deals with the difficulties of native Hungarian learners in oral production. While the sounds of the two languages have some similarities, most of the interferences are to be found in the vowel oppositions (Hungarian phonological duration vs French phonetic duration), the pronunciation of certain consonants and prosody (Hungarian tonic intensity stress vs durational rhythmic French accent). Various experimental methods commonly used at the Institut de Phonétique de Strasbourg were applied : spectrography for the quality and duration of vowels, and for the prosodic characteristics, laryngography for pitch assimilations and consonant contacts. Analyses therefore lean both on the segmental and the supra-segmental levels.This work is more than just a comparative and contrastive study of languages, as it goes beyond by focusing on didactic and pedagogical purposes, using explanations and remediation tips. We hope that it will be useful to French teachers in Hungary, but also to our Hungarian colleagues who may not always be aware of some typical and redundant faults, starting from the same substrate as their learners.

Esta pesquisa tem como objetivo principal investigar os padrões entoacionais na construção de sentido em produções de crianças do 1 ano do ensino fundamental, utilizando diferentes gêneros textuais. Para isso, foram investigadas crianças de um Colégio particular da Cidade do Recife, na faixa etária entre 6 e 7 anos, tanto do gênero masculino quanto do feminino. Os dados foram coletados durante atividades de leitura, onde a professora contou histórias para seus alunos utilizando diferentes gêneros textuais, como fábula, conto de fadas, história em quadrinhos e carta. A pesquisa apresenta uma análise qualitativa da atividade de leitura, analisando o papel da entoação na construção de sentido. A análise dos dados foi realizada com base no modelo Interativo Entoacional de David Brazil (1985), buscando identificar as pistas entoacionais fornecidas na leitura da professora para seus alunos e sua relação com a construção de sentido. Os resultados mostram que a entoação assume três funções textuais: a organizacional, a informativa e a interacional, que apresentam um papel significativo para a construção de sentido durante a atividade de leitura, e que as palavras com proeminências utilizadas pela professora, em sua grande maioria, foram as mesmas escolhidas por seus alunos, contribuindo para construir o sentido no momento da leitura. A pesquisa oferece subsídios tanto para o desenvolvimento e aprimoramento do olhar do fonoaudiólogo escolar com relação à oralidade e aos seus aspectos prosódicos, na produção de crianças em processo de desenvolvimento da aquisição da leitura-escrita, como para o desenvolvimento de estratégias que possam diminuir as dificuldades na produção oral da criança, através da construção de momentos de interação entre terapeuta e paciente, promovendo a elaboração do texto oral da criança, dando um novo sentido a sua leitura

Com o objetivo de contribuir para a caracterização da variedade falada no noroeste paulista, bem como de auxiliar na compreensão da diversidade linguística encontrada no Brasil, descrevemos as características fonético-fonológicas da variedade falada em São José do Rio Preto, utilizando o referecial teórico-metodológico da Sociolingüística Laboviana e da Dialetologia. Analisaremos doze amostras de fala do banco de dados do Projeto Iboruna, selecionadas segundo os fatores sociais: sexo, faixa etária e grau de escolaridade. Descreveremos as vogais orais e nasais, as realizações dos fonemas /s/, /l/, /´/, /r/ e /d/ e /t/ antes de [i] e outros fenômenos linguísticos. Fizemos também referência a obras que tratam da descrição de outras variedades do português com o objetivo de observar se as realizações encontradas em São José do Rio Preto são conservadoras ou inovadoras, se são peculiaridades de lá, ou se também são encontradas em outras variedades. Concluímos que a variedade estudada apresenta características fonético-fonológicas que são comuns em outras comunidades lusófonas e que, se partirmos do português europeu para se estabelecer comparações com a variedade estudada, esta apresenta algumas características conservadoras. / Aiming to contribute to the characterization of the variety spoken in the northwestern region of São Paulo, as well as help understanding the linguistic diversity found in Brazil, the phonetic-phonological features of the variety spoken in São José do Rio Preto were described, using the theoretical and methodological references of the Labovian Sociolinguistics and the Dialectology. Twelve samples of speech from Project Iboruna database, selected according to the following social factors: gender, age and education level will be analyzed. The oral and nasal vowels, the pronunciation of the phonemes /s/, /l/, /´/, /r/ and /d/ and /t/ before [i], and other linguistic phenomena were analyzed. The literature on the description of other varieties of Portuguese language was referred in order to observe if the variants found in São José do Rio Preto are conservative or innovative, if they are peculiarities from there, or whether they are found in other varieties. The conclusion was that the analyzed variety presents phonetic-phonological features that are common in other lusophone communities, and if compared to the European Portuguese, it has some conservative features.