No description available.

Thesis (PhD)--Stellenbosch University, 2014. / ENGLISH ABSTRACT: This thesis finds its epistemological basis in two related motives: the re-conceptualisation of white writing in Zimbabwe as a sub-category of Zimbabwean literature, and the recognition of white narratives as necessarily dialogic. The first motive follows the realization that writing by Zimbabwean whites is systematically marginalized from “mainstream” Zimbabwean literature owing to its perceived irrelevance to the postcolonial Zimbabwean nation. Through an application of Even-Zohar’s polysystem theory, this thesis argues for a recognition of white writing as a literary sub-system existing in relation to other literary and non-literary systems in Zimbabwe’s polysystem of culture. As its second motive, the thesis also calls for a critical approach to white Zimbabwean narratives built on the understanding that the study of literature can no longer be left to monologic approaches alone. Rather, white narratives should be considered as multiple and hence amenable to a multiplicity of approaches that recognize dialogue as an essential aspect of all narratives. The thesis attempts, by closely reading nine white-authored narratives in Zimbabwe, to demonstrate that white Zimbabwean literature is characterized by multiplicity, simultaneity and instability; these are tropes developed from Bakhtin’s understanding of utterances as characterized by a minimum of two voices. To consider white writing in Zimbabwe as a multiplicity is to call forth its numerous dimensions and breadth of perceptions. Simultaneity posits the need to understand opposites/conflicts as capable of existing side by side without necessarily dissolving into unity. Instability captures the several movements and destabilizations that affect writers, characters and the literary system. These three tropes enable a re-reading of white Zimbabwean narratives as complex and multi-nuanced. Such characteristics of the literary system are seen to reflect on the experiences of “whiteness” in postcolonial Zimbabwe. The white narratives selected for examination in this thesis therefore exhibit crises of belonging that reflect the dialogic nature of existence. In sum, this thesis is meant as a dialogue, culminating in the proposition that calls for a decentred and redemptive literary experience. / AFRIKKANSE OPSOMMING: Hierdie tesis vestig sy epistemologiese basis in twee verwante motiewe: die herkonseptualisering van skryfwerk deur wit skrywers in Zimbabwe as ’n sub-kategorie van Zimbabwiese letterkunde, en die erkenning van wit narratiewe as onontkombaar dialogies in aard en wese. Die eerste motief volg die argument dat die skryfwerke van wit Zimbabwieërs stelselmatig gemarginaliseer is uit “hoofstroom” Zimbabwiese literatuur, as gevolg van dié skryfwerke se beweerde irrelevansie tot die koloniale Zimbabwiese nasie-staat. Deur Even-Zohar se polisisteem teorie toe te pas, pleit hierdie tesis vir die erkenning van letterkunde deur wit skrywers as ’n literêre sub-stelsel wat bestaan in verhouding tot ander literêre en nie-literêre sisteme in Zimbabwe se polisisteem van kultuur. As sy tweede motief, vra die tesis ook vir ’n kritiese benadering tot wit Zimbabwiese narratiewe, gebou op die verstandhouding dat die studie van letterkunde nie meer suiwer aan monologies benaderings oorgelewer behoort te word nie. Inteendeel, wit narratiewe moet as veelsydig beskou word, en dus vatbaar vir ’n verskeidenheid benaderings wat dialoog as ’n noodsaaklike aspek van alle verhale erken en verken. Deur nege wit outeurs se verhale in Zimbabwe noukeurig te lees, dui hierdie tesis aan dat wit Zimbabwiese literatuur gekenmerk word deur veelvuldigheid, gelyktydigheid en onstabiliteit; hieride is teoretiese konsepte wat ontleen is aan Bakhtin se begrip van uitsprake (“utterances”) as bestaande uit ’n minimum van twee stemme. Om wit lettere in Zimbabwe as veelvuldig te verklaar is om die talle dimensies en breedtes van persepsie in letterkundige korpus te erken. Gelyktydig postuleer die tesis die moontlikheid dat teenoorgesteldes/konflikte langs mekaar kan en móét bestaan, sonder om noodwendig in ’n eenheid te ontaard. Onstabiliteit, soos dit hier verstaan word, omvat die verskillende bewegings en ontstuimige roeringe wat skrywers, karakters en die literêre sisteem beïnvloed. Hierdie drie konsepte laat ’n herlees van wit Zimbabwiese verhale toe wat as kompleks en multi-genuanseerd bestempel kan word. Sulke kenmerke van die literêre sisteem moet in ag geneem word om die ervaring van “witheid” in post-koloniale Zimbabwe effektief uit te beeld. Die wit verhale wat gekies is vir herlees in hierdie tesis beeld dus krisisse van bestaan uit wat die dialogiese aard van die menslike bestaan omvat. Ter afsluiting is hierdie tesis bedoel as ’n dialoog wat kulmineer in ’n oproep vir gedensentraliseerde en verlossende ervarings van die letterkunde in sy geheel.

L’objectif de cette thèse est de concevoir et de construire, un système Text-To-Speech (TTS) haute qualité à base de HMM (Hidden Markov Model) pour le vietnamien, une langue tonale. Le système est appelé VTED (Vietnamese TExt-to-speech Development system). Au vu de la grande importance de tons lexicaux, un tonophone” – un allophones dans un contexte tonal – a été proposé comme nouvelle unité de la parole dans notre système de TTS. Un nouveau corpus d’entraînement, VDTS (Vietnamese Di-Tonophone Speech corpus), a été conçu à partir d’un grand texte brut pour une couverture de 100% de di-phones tonalisés (di-tonophones) en utilisant l’algorithme glouton. Un total d’environ 4000 phrases ont été enregistrées et pré-traitées comme corpus d’apprentissage de VTED.Dans la synthèse de la parole sur la base de HMM, bien que la durée de pause puisse être modélisée comme un phonème, l’apparition de pauses ne peut pas être prédite par HMM. Les niveaux de phrasé ne peuvent pas être complètement modélisés avec des caractéristiques de base. Cette recherche vise à obtenir un découpage automatique en groupes intonatifs au moyen des seuls indices de durée. Des blocs syntaxiques constitués de phrases syntaxiques avec un nombre borné de syllabes (n), ont été proposés pour prévoir allongement final (n = 6) et pause apparente (n = 10). Des améliorations pour allongement final ont été effectuées par des stratégies de regroupement des blocs syntaxiques simples. La qualité du modèle prédictive J48-arbre-décision pour l’apparence de pause à l’aide de blocs syntaxiques, combinée avec lien syntaxique et POS (Part-Of-Speech) dispose atteint un F-score de 81,4 % (Précision = 87,6 %, Recall = 75,9 %), beaucoup mieux que le modèle avec seulement POS (F-score=43,6%) ou un lien syntaxique (F-score=52,6%).L’architecture du système a été proposée sur la base de l’architecture HTS avec une extension d’une partie traitement du langage naturel pour le Vietnamien. L’apparence de pause a été prédit par le modèle proposé. Les caractéristiques contextuelles incluent les caractéristiques d’identité de “tonophones”, les caractéristiques de localisation, les caractéristiques liées à la tonalité, et les caractéristiques prosodiques (POS, allongement final, niveaux de rupture). Mary TTS a été choisi comme plateforme pour la mise en oeuvre de VTED. Dans le test MOS (Mean Opinion Score), le premier VTED, appris avec les anciens corpus et des fonctions de base, était plutôt bonne, 0,81 (sur une échelle MOS 5 points) plus élevé que le précédent système – HoaSung (lequel utilise la sélection de l’unité non-uniforme avec le même corpus) ; mais toujours 1,2-1,5 point de moins que le discours naturel. La qualité finale de VTED, avec le nouveau corpus et le modèle de phrasé prosodique, progresse d’environ 1,04 par rapport au premier VTED, et son écart avec le langage naturel a été nettement réduit. Dans le test d’intelligibilité, le VTED final a reçu un bon taux élevé de 95,4%, seulement 2,6% de moins que le discours naturel, et 18% plus élevé que le premier. Le taux d’erreur du premier VTED dans le test d’intelligibilité générale avec le carré latin test d’environ 6-12% plus élevé que le langage naturel selon des niveaux de syllabe, de ton ou par phonème. Le résultat final ne s’écarte de la parole naturelle que de 0,4-1,4%. / The thesis objective is to design and build a high quality Hidden Markov Model (HMM-)based Text-To-Speech (TTS) system for Vietnamese – a tonal language. The system is called VTED (Vietnamese TExt-tospeech Development system). In view of the great importance of lexical tones, a “tonophone” – an allophone in tonal context – was proposed as a new speech unit in our TTS system. A new training corpus, VDTS (Vietnamese Di-Tonophone Speech corpus), was designed for 100% coverage of di-phones in tonal contexts (i.e. di-tonophones) using the greedy algorithm from a huge raw text. A total of about 4,000 sentences of VDTS were recorded and pre-processed as a training corpus of VTED.In the HMM-based speech synthesis, although pause duration can be modeled as a phoneme, the appearanceof pauses cannot be predicted by HMMs. Lower phrasing levels above words may not be completely modeled with basic features. This research aimed at automatic prosodic phrasing for Vietnamese TTS using durational clues alone as it appeared too difficult to disentangle intonation from lexical tones. Syntactic blocks, i.e. syntactic phrases with a bounded number of syllables (n), were proposed for predicting final lengthening (n = 6) and pause appearance (n = 10). Improvements for final lengthening were done by some strategies of grouping single syntactic blocks. The quality of the predictive J48-decision-tree model for pause appearance using syntactic blocks combining with syntactic link and POS (Part-Of-Speech) features reached F-score of 81.4% Precision=87.6%, Recall=75.9%), much better than that of the model with only POS (F-score=43.6%)or syntactic link (F-score=52.6%) alone.The architecture of the system was proposed on the basis of the core architecture of HTS with an extension of a Natural Language Processing part for Vietnamese. Pause appearance was predicted by the proposed model. Contextual feature set included phone identity features, locational features, tone-related features, and prosodic features (i.e. POS, final lengthening, break levels). Mary TTS was chosen as a platform for implementing VTED. In the MOS (Mean Opinion Score) test, the first VTED, trained with the old corpus and basic features, was rather good, 0.81 (on a 5 point MOS scale) higher than the previous system – HoaSung (using the non-uniform unit selection with the same training corpus); but still 1.2-1.5 point lower than the natural speech. The quality of the final VTED, trained with the new corpus and prosodic phrasing model, progressed by about 1.04 compared to the first VTED, and its gap with the natural speech was much lessened. In the tone intelligibility test, the final VTED received a high correct rate of 95.4%, only 2.6% lower than the natural speech, and 18% higher than the initial one. The error rate of the first VTED in the intelligibility test with the Latin square design was about 6-12% higher than the natural speech depending on syllable, tone or phone levels. The final one diverged about only 0.4-1.4% from the natural speech.

O objetivo geral desta dissertação é fornecer um quadro amplo da formação dos compostos nas línguas naturais, deflagrando o que deve ser visto como universal em sua formação e o que deve ser visto como particular às línguas. Partindo de uma língua específica, o português brasileiro, demonstramos como a composição se estabelece em um sistema particular, e, no contraste entre as propriedades de seus dados e as propriedades atestadas nesse processo em trabalhos taxonômicos e tipológicos, delineamos as fronteiras entre o geral e o específico. Admitimos que sejam duas as propriedades universais da composição que devem ser abarcadas em um sistema gerativo: (i) o estabelecimento de uma relação gramatical entre os membros de um composto, a saber, uma relação de subordinação, atribuição ou coordenação fato bem comprovado translinguisticamente nos trabalhos de Bisetto e Scalise (2005), Guevara e Scalise (2009), e Scalise e Bisetto (2009) , e (ii) a criação de um domínio categorial acima de dois núcleos complexos conectados em determinada relação gramatical, o qual garante que uma estrutura complexa quase-sentencial seja interpretada como uma única unidade sintática. O particular nesse processo de formação de palavras reside no modo como as línguas naturais emolduram morfologicamente seus compostos, ou se a partir de uma combinação entre radicais, ou se a partir de uma combinação entre palavras, bem como os processos fonológicos que determinam a composição em um sistema linguístico específico. Essa assunção é evidenciada pela variedade de estruturas morfológicas encontrada nos compostos translinguisticamente, e pela assistematicidade de processos fonológicos que se aplicam nos compostos a fim de demarcá-los em uma língua. Para explicar esses fatos, assumimos um modelo não-lexicalista de gramática, a Morfologia Distribuída (cf. HALLE; MARANTZ, 1993; EMBICK; NOYER, 2007), e, com base nesse aparato teórico, demonstramos que as propriedades universais da composição são diretamente abarcadas no componente sintático. As relações gramaticais, primeiramente, são formalizadas através dos tipos de aplicação da operação Merge, tal como definidos em Chomsky (2000, 2004), nomeadamente, set-Merge e pair-Merge, enquanto que o domínio categorial é formado a partir da concatenação de um núcleo definidor de categoria acima de duas ou mais raízes categorizadas. As estruturas morfológicas, por sua vez, serão emolduradas pós-sintaticamente, nos componentes morfológico e fonológico da gramática, onde a variação translinguística se processa. Para tanto, argumentamos que a informação de classe carrega um papel importante na variação estrutural dos compostos translinguisticamente, pois será a presença ou ausência dessa informação que determinará quando uma raiz sintática será um radical ou uma palavra, distribuindo, desse modo, as estruturas sintáticas dos compostos em combinações morfológicas variadas / The main goal of this thesis is to provide a broad picture of compounding in natural languages, triggering what should be considered universal and what should be considered languagespecific in this word formation process. Starting from a particular language, Brazilian Portuguese, we demonstrate how compounding is established in a linguistic system, and by the contrast of the properties present in its compounds with the attested properties provided in taxonomic and typological works on the subject, we outline the boundaries between what is general and what is specific to this language. We argue that there are two universal properties of compounding that should be accounted for a generative system: (i) the establishment of a grammatical relation between the constituents of a compound, namely subordination, attribution and coordination a crosslinguistically well-proven fact in the works of Bisetto and Scalise (2005), Guevara and Scalise (2009) and Scalise and Bisetto (2009) and (ii) the formation of a categorial domain above two constituents connected in a specific grammatical relation, which ensures that a quasi-sentential structure will be interpreted as a single syntactic unit. What is language-specific in this word formation process lies in how languages frame their compounds in morphological structures, whether in a combination of stems or in a combination of words, as well as the phonological processes that specify compounding in some linguistic systems. This assumption finds evidence in the variety of morphological structures found in compounds through the languages of the world, and in the unsystematicity of phonological processes that apply to compounds to demarcate it in a particular language. In order to explain all the above-mentioned facts, we assume a non-lexicalist approach to grammar, the Distributed Morphology framework (see HALLE; MARANTZ, 1993; EMBICK; NOYER, 2007), and based on its tenets we demonstrate that the universal properties of compounding are straightforwardly accounted for the syntactic component of the grammar. Primarily, the grammatical relations are formalized through the nature of the operation Merge, as defined in Chomsky (2000, 2004), namely, set-Merge and pair-Merge, while the categorial domain is created by the concatenation of a category-defining head above two or more categorized roots. The morphological structures of the compounds will be framed post-syntactically at the morphological and phonological components, where crosslinguistic variation takes place. To this end, we argue that class marker information plays an important role in the variation of compounds morphological structure crosslinguistically, since it will be the absence or the presence of this feature that will define whether a syntactic root will be externalized as a stem or as a word, distributing this way the syntactic structure of a compound in various morphological combinations

Esta dissertação visa analisar o uso no português brasileiro, doravante PB, da pista prosódica de duração de sílabas na produção e percepção de sentenças ambíguas do tipo SN1-V-SN2-Atributo, tais como A mãe encontrou a filha suada. Tais sentenças apresentam contextos em que pode haver reestruturação de frases fonológicas a depender da interpretação escolhida. Um dos trabalhos que guiaram esta pesquisa foi o de MAGALHÃES & MAIA (2006), no qual os autores também observaram sentenças com ambiguidade devido à presença de um atributo que pode se referir ao sujeito ou ao objeto da oração. Fonologicamente, tais leituras podem ser explicadas pelo fato de o atributo poder ou não se juntar a seu núcleo na construção do domínio da frase fonológica, em que, se há fronteira, um alongamento é esperado (NESPOR & VOGEL, 1996). ANGELO & SANTOS (2012, 2015) concluíram que os falantes não realizaram alongamento significativo, porém, observaram uma tendência a produzir sentenças de aposição não local mais longas. O que se propõe, então, é que o alongamento é um fenômeno existente na língua, mas é opcional. Se isto é verdade, sempre que ele for feito, a sentença deve ser interpretada como não local. Caso contrário, haveria uma variação nas respostas perceptuais ou preferência por aposição local (Princípio da Aposição Local verificado por MAGALHÃES & MAIA (2006) no PB com base no Princípio de Late Closure apontado por FRAZIER (1979)). Dois testes foram realizados: No de produção, além dos 30 falantes de Angelo & Santos, mais 20 foram escolhidos e outros 10 recuperados para que lessem ambas as versões de cada uma das 9 sentenças em meio a histórias que guiavam a um ou outro significado, totalizando 720 dados. No de percepção, foram selecionadas as 3 versões mais longas (de aposição não local) e as 3 mais curtas (de aposição local) do teste de produção para cada sentença, em um teste onde 30 ouvintes selecionaram através de imagens qual a interpretação obtida, totalizando 1620 dados. Primeiramente, os resultados apontaram para diferenças significativas observando o tipo de estrutura (os falantes alongaram e acertaram as não locais). Por tipo de sentença, a diferença foi significativa na maioria delas, confirmando a predição de que, ainda que o alongamento seja opcional na produção, uma vez realizado, ele serve como condutor para uma interpretação não local. Para as sentenças locais, o Princípio de Aposição Local verificado no PB por MAGALHÃES & MAIA (2006) também foi observado em nosso experimento. Por fim, os resultados apontam para 2 tipos de estruturas que podem interferir no processo de alongamento: Sentenças em que o atributo é formado por adjetivos deverbais (geração de uma fronteira CP) e sentenças que permitem construções de small clause (bloqueio da reestruturação). / The aim of this thesis is to analyze the use of syllable duration in Brazilian Portuguese (henceforth BP) as a prosodic cue in the perception and production of syntactically ambiguous sentences with a NP1-V-NP2-attribute structure (lit. The mother has found her daughter sweating). This type of sentence is produced in contexts where phonological phrases may be restructured, depending on how the sentences are interpreted. A reference for the present research was the study by MAGALHÃES & MAIA (2006), who also found sentences which ambiguity was due to the presence of an attribute that may refer to either the subject or the object of the clause. Phonologically, interpretations of the above-mentioned structure can be explained by the fact that attributes may or may not join the head in the construction of the phonological phrase domain, because lengthening is expected when there is a boundary (NESPOR & VOGEL, 1996). ANGELO & SANTOS (2015) found that the speakers in their study did not make significant lengthening; however, they tended to produce high attachment in longer sentences. It is suggested, therefore, that although lengthening exists in BP as a phenomenon, it is optional. If this is actually true, sentences should be interpreted as having high attachment whenever lengthening is produced. Otherwise, it is assumed that there is variation in perceptual responses or preference for low attachment (Low Attachment Principle, found in PB by MAGALHÃES & MAIA (2006), based on the Late Closure Principle proposed by Frazier (1979)). Data were collected with two tests. In the sentence-reading production test, 30 speakers from the study by ANGELO & SANTOS (2015)) and 20 new speakers read one version of each sentence, while 10 out of the former 30 read two versions, thus the corpus was made of a total of 720 sentences. The perception test used the three longest versions (with high attachment) and the three shortest versions (with low attachment) of each sentence. The informants who took the test chose the image that best represented the meaning that they assigned to each sentence. A total of 1620 data were produced. Overall results showed significant differences for type of structure (the informants lengthened and identified high attachments correctly). For sentence type, there were significant differences in most sentences, thus confirming the hypothesis that lengthening is optional in production and, once it has been performed, it can lead a sentence to be interpreted as having high attachment. For sentences with low attachment, the Low Attachment Principle suggested by MAGALHÃES & MAIA (2006) also occurred in our experiment. The findings signal two types of structures that may interfere with the lengthening process: sentences whose attribute is formed by deverbal adjectives (generation of a CP phrasal boundary) and sentences that allow small clause constructions (restructuring is blocked).

Theories of phonological word formation (e.g. Selkirk 1980, 1986; Nespor & Vogel 1986; Lahiri & Plank 2010) assume that prosodic units are not isomorphic with syntactic units. However, the prosodic status of compounds remain uncertain, at least in so far as language planning and phonological encoding is concerned. Theories are not transparent about the prosodic status of compounds: although a noun-noun compound in English consists of two lexical words (and therefore two prosodic words), it can also act as a single prosodic item by exhibiting main stress on the first unit and carrying inflection. Thus the question remains controversial - should these items be treated as a single prosodic unit, similar to a monomorphemic word, or as two distinct units for the purpose of post-lexical representation? Recursive word formation may suggest that compounds are a single unit. Psycholinguistic evidence measuring speech onset latency in native speakers of Dutch and Portuguese also shows compounds being treated as single prosodic units (Wheeldon & Lahiri 1997, 2002; Vigario, 2010). Although recent studies have produced evidence for the prosodification of compounds in native speakers, little is known about the process in non-native speakers. Our research questions are as follows: what is the post-lexical planning unit in English, and how do non-native fluent speakers of English plan these units for the purpose of phonological encoding? To investigate our hypotheses, we focus on the phonological encoding of compounds with and without encliticisation, for native and non-native speakers of English. In a series of delayed priming tasks, we found overwhelming evidence that reaction times reflected the total number of prosodic units in the target sentence. In online tasks, however, speech latencies only reflected the size of the first prosodic unit. Taken together,these results suggest that, despite containing two lexical and prosodic words, English compounds are planned as single prosodic units, exhibiting encliticisation and reaction times similar to those of monomorphemic words. As shown by the results in this study, this naming paradigm has proved extremely beneficial for eliciting data about the structure of prosodic units in speech production. Not only was it successful for native speakers of Dutch, European Portuguese, and English, we also found that it was easily implemented into a study of post-lexical encoding in non-native speakers of English.

Prosody is the melody and rhythm of speech and language. Prosodic ability is of great importance for communication, and has been shown to be affected in adults with acquired brain injury. Studies have shown that Swedish children with specific language impairment have prosodic  difficulties to a greater extent than e.g. English speaking peers. It is thereby possible that prosodic ability is more vulnerable also in Swedish speaking adults with acquired brain injury. The aim of the present study was to assess prosodic ability in adults with acquired brain injury, and to evaluate the prosodic assessment procedure that was used. The relationship between prosodic ability and general language ability, location of injury and post-onset time was examined. Fifteen adults with acquired brain injury participated, six women and nine men, aged between 30 and 82 years. The participants were tested with an assessment procedure for prosody and parts of Neurolinguistic Assessment of Aphasia (A-ning). Prosodic problems were found at word, phrase and discourse level. No clear links between prosodic ability and location of injury or time post-onset were found. However, a deviant usage of focal accents with a preserved ability to produce word accents as a result of right hemisphere injuries was found. Furthermore, a relationship between prosodic ability and general language ability was found. The prosodic procedure was proven useful for assessment of adults with acquired brain injury, but it needs additional revision. The present study contributes with additional knowledge of prosodic ability in native Swedish speaking adults with acquired brain injury. / Prosodi är talets och språkets melodi och rytm. Den prosodiska förmågan är av stor betydelse för kommunikationen, och är ofta påverkad hos vuxna personer med förvärvad hjärnskada. Forskning har visat att svenska barn med språkstörning har prosodiska svårigheter i större utsträckning än exempelvis engelsktalande barn med språkstörning. Därför är det möjligt att den prosodiska förmågan är mer sårbar även hos vuxna personer med förvärvad hjärnskada och svenska som modersmål. Syftet med föreliggande studie var att kartlägga prosodisk förmåga hos vuxna personer med förvärvad hjärnskada, samt att utvärdera det undersökningsmaterial för prosodi som användes. Sambandet mellan prosodisk förmåga och språkförmåga, skadelokalisation och tid sedan skada undersöktes. I studien deltog 15 vuxna personer, 6 kvinnor och 9 män, mellan 30 och 82 år med förvärvad hjärnskada. Deltagarna testades avseende prosodi samt med delar ur Neurolingvistisk afasiundersökning (A-ning). Resultatet visade att prosodiska avvikelser förekom hos deltagarna på ord- fras- och diskursnivå. Inga tydliga samband fanns mellan prosodisk förmåga och skadelokalisation eller tid sedan skada. Högersidiga skador föreföll dock ge svårigheter att använda fokal accent, trots bevarad förmåga att använda ordaccenter. Vidare sågs ett samband mellan prosodisk och språklig förmåga. Undersökningsmaterialet för prosodi var användbart för testning av personer med förvärvad hjärnskada, men behöver utarbetas ytterligare. Föreliggande studie bidrar till kunskapen om prosodisk förmåga hos vuxna personer med förvärvad hjärnskada och svenska som modersmål.

Current high quality text-to-speech (TTS) systems are based on unit selection from a large database that is both contextually and prosodically rich. These systems, albeit capable of natural voice quality, are computationally expensive and require a very large footprint. Their success is attributed to the dramatic reduction of storage costs in recent times. However, for many TTS applications a smaller footprint is becoming a standard requirement. This thesis presents a new method for representing speech segments that can improve the quality and/or reduce the footprint current concatenative TTS systems. The circular linear prediction (CLP) model is revisited and combined with the constant pitch transform (CPT) to provide a robust representation of speech signals that allows for limited prosodic movements without a perceivable loss in quality. The CLP model assumes that each frame of voiced speech is an infinitely periodic signal. This assumption allows for LPC modeling using the covariance method, with the efficiency of the autocorrelation method. The CPT is combined with this model to provide a database that is uniform in pitch for matching the target prosody during synthesis. With this representation, limited prosody modifications and unit concatenation can be performed without causing audible artifacts. For resolving artifacts caused by pitch modifications in voicing transitions, a method has been introduced for reducing peakiness in the LP spectra by constraining the line spectral frequencies. Two experiments have been conducted to demonstrate the potential for the capabilities of CLP/CPT method. The first is a listening test to determine the ability of this model to realize prosody modifications without perceivable degradation. Utterances are resynthesized using the CLP/CPT method with emphasized prosodics to increase intelligibility in harsh environments. The second experiment compares the quality of utterances synthesized by unit-selection based limited-domain TTS against the CLP/CPT method. The results demonstrate that the CLP/CPT representation, applied to current concatenative TTS systems, can reduce the size of the database and increase the prosodic richness without noticeable degradation in voice quality.

This thesis develops the phonological model for the Stockholm Swedish intonation system. Though previous research provides a general model of this system, many phonological aspects of it have remained understudied. The intonational options that are available to speakers of Stockholm Swedish are discussed, and it is argued that Stockholm Swedish provides evidence for complex branching of phonological domains. Specifically, it is argued that so called focal accents, which are referred to as (H)LH-accents in the present work, have essentially two different functions. First, they signal information structural categories such as focus. Second, they signal left edges of Intonation Phrases (IP). It is also argued that a wide range of options exist in the post-nuclear area. Six types of contours for such areas are distinguished, plus one additional rising contour when there are no post-nuclear accents. Based on these findings, I present an account of the branching options for the phonological categories in the Stockholm Swedish prosodic hierarchy. I argue that there is evidence for recursive phonological structures in Stockholm Swedish, i.e. that a mother node and a daughter node can belong to the same phonological category. Also, Stockholm Swedish provides evidence for a distinction between prosodic coordination (equal sister nodes) and prosodic adjunction (unequal sister nodes). Prosodic structure is mapped onto syntactic structure via a set of variably ranked Optimality Theoretic constraints. The relation between phonological and syntactic structure shows that the phonology prefers prosodic coordination (equal sisters) over adjunction (unequal sisters). The material for the study comprises a corpus of approximately 420 read sentences, which were specifically designed to test various phonological hypotheses, and approximately 17 minutes of uncontrolled speech.

xviii, 216 p. : ill. (some col.) / The focus of this dissertation is on how language experience shapes perception of a non-native prosodic contrast. In Tokyo Japanese, fundamental frequency (F0) peak and fall are acoustic cues to lexically contrastive pitch patterns, in which a word may be accented on a particular syllable or unaccented (e.g., tsúru 'a crane', tsurú 'a vine', tsuru 'to fish'). In English, lexical stress is obligatory, and it may be reinforced by F0 in higher-level prosodic groupings. Here I investigate whether English listeners can attend to F0 peaks as well as falls in contrastive pitch patterns and whether training can facilitate the learning of prosodic categories. In a series of categorization and discrimination experiments, where F0 peak and fall were manipulated in one-word utterances, the judgments of prominence by naïve English listeners and native Japanese listeners were compared. The results indicated that while English listeners had phonetic sensitivity to F0 fall in a same-different discrimination task, they could not consistently use the F0 fall to categorize F0 patterns. The effects of F0 peak location and F0 fall on prominence judgments were always larger for Japanese listeners than for English listeners. Furthermore, the interaction between these acoustic cues affected perception of the contrast by Japanese, but not English, listeners. This result suggests that native, but not non-native, listeners have complex and integrated processing of these cues. The training experiment assessed improvement in categorization of Japanese pitch patterns with exposure and feedback. The results suggested that training improved identification of the accented patterns, which also generalized to new words and new contexts. Identification of the unaccented pattern, on the other hand, showed no improvement. Error analysis indicated that native English listeners did not learn to attend specifically to the lack of the F0 fall. To conclude, language experience influences perception of prosodic categories. Although there is some sensitivity to F0 fall in non-native listeners, they rely mostly on F0 peak location in language-like tasks such as categorization of pitch patterns. Learning of new prosodic categories is possible. However, not all categories are learned equally well, which suggests that first language attentional biases affect second language acquisition in the prosodic domain. / Committee in charge: Susan Guion Anderson, Chairperson; Melissa A. Redford, Member; Vsevolod Kapatsinki, Member; Kaori Idemaru, Outside Member