京都大学 / 新制・課程博士 / 博士(情報学) / 甲第24256号 / 情博第800号 / 新制||情||135(附属図書館) / 京都大学大学院情報学研究科知能情報学専攻 / (主査)教授 河原 達也, 教授 森 信介, 教授 伊藤 孝行 / 学位規則第4条第1項該当 / Doctor of Informatics / Kyoto University / DFAM

BACKGROUND: Many families of children who are deaf or hard of hearing (DHH) pursue the development of spoken language through the use of advanced hearing technology and intensive, specialized listening and spoken language (LSL) intervention services. LSL practitioners utilize caregiver coaching to transfer knowledge and skills to parents, equipping them to effectively support their child's listening and language development. Caregiver coaching builds parents' capacity to implement intervention strategies within daily routines to maximize learning opportunities to reach this goal. While caregiver coaching is a hallmark of LSL practice, there is a lack of consensus and paucity of evidence to support its use with children who are DHH learning to listen and talk. The purpose of this research was therefore to gain a greater understanding of caregiver coaching practices in LSL services for families of children ages birth-3 who are DHH. OBJECTIVES: 1) to provide a synthesis of the literature and identify gaps in the existing knowledge base regarding coaching in LSL services, 2) to gain insight into practitioners' and caregivers' experiences with coaching in LSL services, and 3) to gain a better understanding of coaching practices in LSL services with families of children who are DHH. METHODOLOGY: The first objective was addressed by performing a scoping review of the literature to synthesize the relevant research and professional practice recommendations and identify gaps in the knowledge base regarding coaching caregivers in LSL practice. The remaining objectives were addressed through qualitative, semi-structured interviews and video observation discussions with LSL practitioners and caregivers to gain insight into caregivers' perspectives and practitioners' perspectives and practices. Interviews were conducted with 14 practitioners and 13 caregivers at three intervention sites in the US and Canada. These sites represent three different service delivery models of LSL intervention for families of children who are DHH. RESULTS: The scoping review findings revealed a lack of consensus in the literature regarding the principles and practices of caregiver coaching in LSL services. We presented the following topics found in the literature: coaching practices, training for coaching, the effectiveness of coaching, and recommendations for coaching. Caregivers reported coaching as a positive experience; however, coaching practices differed among sites and between practitioners, supporting the results of the scoping review. The caregivers indicated factors that contribute to a positive coaching relationship, including practitioner characteristics, establishing explicit expectations, and adapting to caregivers' changing needs over time. The practitioner interviews also supported a lack of consistency in coaching practices between sites and indicated that underlying beliefs impact how practitioners coach and engage caregivers. CONCLUSION: The findings from this dissertation provide practical, actionable steps that LSL practitioners can implement to develop and support the caregiver coaching relationship. These findings have the potential to inform professional preparation and development activities to better equip practitioners to engage caregivers in the intervention process, and ultimately, positively impact the listening and spoken language outcomes of the children they serve.

Sequence-to-sequence learning led to significant improvements to machine translation (MT) and automatic speech recognition (ASR) systems. These advancements were first reflected in spoken language translation (SLT) when using a cascade of (at least) ASR and MT with the new "neural" models, then by using sequence-to-sequence learning to directly translate the input audio speech into text in the target language. In this thesis we cover both approaches to the SLT task. First, we show the limits of NMT in terms of robustness to input errors when compared to the previous phrase-based state of the art. We then focus on the NMT component to achieve better translation quality with higher computational efficiency by using a network based on weakly-recurrent units. Our last work involving a cascade explores the effects on the NMT robustness when adding automatic transcripts to the training data. In order to move to the direct speech-to-text approach, we introduce MuST-C, the largest multilingual SLT corpus for training direct translation systems. MuST-C increases significantly the size of publicly available data for this task as well as their language coverage. With such availability of data, we adapted the Transformer architecture to the SLT task for its computational efficiency . Our adaptation, which we call S-Transformer, is meant to better model the audio input, and with it we set a new state of the art for MuST-C. Building on these positive results, we finally use S-Transformer with different data applications: i) one-to-many multilingual translation by training it on MuST-C; ii participation to the IWSLT 19 shared task with data augmentation; and iii) instance-based adaptation for using the training data at test time. The results in this thesis show a steady quality improvement in direct SLT. Our hope is that the presented resources and technological solutions will increase its adoption in the near future, so to make multilingual information access easier in a globalized world.

情報処理学会研究報告. SLP, 音声言語情報処理; 2003-SLP-49-24 第5回音声言語シンポジウム

Language Identification (LID) is the task of automatically identifying the language of speech signal uttered by an unknown speaker. An N language LID task is to classify an input speech utterance, spoken by an unknown speaker and of unknown text, as belonging to one of the N languages L1, L2, . . , LN. We present a new approach to spoken language modeling for language identification using the Lempel-Ziv-Welch (LZW) algorithm, with which we try to overcome the limitations of n-gram stochastic models by automatically identifying the valid set of variable length patterns from the training data. However, since several patterns in a language pattern table are also shared by other language pattern tables, confusability prevailed in the LID task. To overcome this, three pruning techniques are proposed to make these pattern tables more language specific. For LID with limited training data, we present another language modeling technique, which compensates for language specific patterns missing in the language specific LZW pattern table. We develop two new discriminative measures for LID based on the LZW algorithm, viz., (i) Compression Ratio Score (LZW-CRS) and (ii) Weighted Discriminant Score (LZW-WDS). It is shown that for a 6-language LID task of the OGI-TS database, the new model (LZW-WDS) significantly outperforms the conventional bigram approach. With regard to the front end of the LID system, we develop a modified technique to model for Acoustic Sub-Word Units (ASWU) and explore its effectiveness. The segmentation of speech signal is done using an acoustic criterion (ML-segmentation). However, we believe that consistency and discriminability among speech units is the key issue for the success of ASWU based speech processing. We develop a new procedure for clustering and modeling the segments using sub-word GMMs. Because of the flexibility in choosing the labels for the sub-word units, we do an iterative re-clustering and modeling of the segments. Using a consistency measure of labeling the acoustic segments, the convergence of iterations is demonstrated. We show that the performance of new ASWU based front-end and the new LZW based back-end for LID outperforms the earlier reported PSWR based LID.

The subject of the master thesis is language attitudes. The focus is put on how language attitudes can affect the way how the language is changing, how they can be expressed and which factors play important roles in forming of language attitudes. There have been a number of sociological and sociolinguistic case studies in this field in Norway. Based on them, I compare the Norwegian and Czech attitudes to the latest trends in language change such as globalization, regionalization and others. The main questions are: how do factors forming language attitudes to language varieties differ when compared Norwegian and Czech? How are language attitudes affected by differencies in history and culture in these two lands? Hypothetical, the factors connected to contact with English would be quite similar both in Norwegian and Czech. On the other hand the relationship between standard spoken language and dialects would differ more mostly because of these varieties have a different status in Norwegian and Czech context.

Research conducted on the prevalence of behavioural adjustment in Deaf children and adolescents, in erstwhile countries, points towards an appreciably elevated percentage of emotional and behavioural problems amongst this population group when compared to hearing normative groups. Studies specify that the prevalence of behaviour and emotional problems in Deaf children and adolescents varies from 4.8% to 50.3%. From existing research conducted, it is ambiguous as to why the reported prevalence rates of maladjustment are higher amongst Deaf children and adolescents. This pioneering study is the first of its kind to research dissimilar modalities of language used as the language of learning and teaching (LoLT) in schools for Deaf learners and how this could possibly correlate to learner behaviour in the classroom. Taking into consideration the reported pervasiveness of maladjustment in Deaf children and adolescents; this study uses the Teacher Report Form (TRF) to investigate the types of behaviour problems displayed by Deaf learners in the classroom. It further investigates whether Deaf learners display certain types of behaviour problems when dissimilar modalities of language are used as the language of learning and teaching. The overall findings of this study suggest that teachers who use manually coded spoken language report an elevated prevalence of behaviour problems on the TRF compared to teachers who use South African Sign Language (SASL). Results further suggest that the group of teachers who use SASL report somatic complaints and attention problems as the most frequently encountered behaviour problems in their classrooms. In comparison the group of teachers who use manually coded spoken English (MCE) report social problems and attention problems as the most frequently encountered behaviour problems in their classrooms. Limitations of this study and suggestions for future research are discussed.

Cette thèse de doctorat a pour objet l’identification et l’analyse des marques d’oralité présentes dans la presse italienne sur Internet à partir de l’exemple de corriere.it, version électronique du quotidien italien publié à Milan, Il Corriere della Sera. Notre étude se fonde sur deux corpus recueillis sur le site du quotidien, l’un de janvier à juin 2004, l’autre de mars à août 2011. Dans les 1356 énoncés que nous avons sélectionnés à parité à travers ces deux corpus, nous avons repéré les phénomènes que nous avons identifiés comme « traits d’oralité» précédemment reconnus comme tels par divers grammairiens et chercheurs. Nous avons classé ces traits d’oralité aux différents niveaux de l’analyse linguistique : la textualité, la morphologie, la syntaxe et le lexique, domaines auxquels correspondent respectivement les quatre chapitres de notre thèse. La méthode d’analyse, quantitative et qualitative, est illustrée par de nombreux graphiques qui visualisent le taux de fréquence des phénomènes, mais nous avons pris aussi en considération certains aspects linguistiques moins représentés dans les deux corpus. Dans la conclusion, nous procédons à une synthèse interprétative des principaux résultats obtenus et nous avançons quelques réflexions sur les aspects sociolinguistiques de la répartition des marques d’oralité et de leur place dans l’évolution de la langue. / The focus of this dissertation is to identify and analyze the traces of orality in the electronic Italian press based on the example of corriere.it, the electronic version of the daily Italian newspaper published in Milan, Il Corriere della Sera. The research project is based on two corpuses assembled from the site of the daily newspaper, one from January through June2004, and the other from March through August 2011. Though the 678 statements selected from each corpus, we highlighted the phenomena which we identified as ‘features of orality”previously recognized as such by various grammarians and scholars. We have classified these features of orality at various levels of linguistic analysis : textuality, morphology, syntax andlexis, fields which correspond to the four chapters of the dissertation respectively. The quantitative and qualitative analytical method is illustrated by numerous graphs whichvisualize the frequency rate of the phenomena, but we also took into consideration certainlinguistic aspects that are sparingly present in the two corpuses. In the conclusion, we drew upan interpretative summary of the main findings and we foregrounded certain reflections on the sociolinguistic aspects of the distribution of the traces of orality and their place in the evolution of languages.

Kalba – pagrindinis žmonių bendravimo rodiklis. Radijas ir televizija - masiškiausios viešosios informacijos priemonės, todėl labai svarbu, kad iš eterio sklindantis žodis būtų gyvas, taisyklingas, kad tvirtintų įsigalėjusias kalbos normas, o ne jas griautų. Šio magistro darbo tikslas - ištirti įvairių žmonių sakytinėje kalboje vartojamas įvardžiuotines formas ir jų funkcijas. Darbe iškelta hipotezė – sakytinėje kalboje vartojamos įvardžiuotinės formos yra retai ar visai nebevartojamos, dažniausiai skirtos tik sudėtiniams terminams ir pavadinimams apibrėžti. Į ją atsakyta pasiremiant pasirinktomis televizijos ir radijo laidomis ir suformuluotomis išvadomis. Tiriamąją medžiagą sudaro devyniolika laidų: septynios radijo laidos (Pozicija, Prie pietų stalo, Ryto garsai, Žmonės ir idėjos, Dienos tema, Druska) ir dvylika televizijos laidų (Įžvalgos, Ką manai, Kultūros namai, Spaudos klubas, Požiūris, Nepatogūs pokalbiai). Darbe pateikiama išsami įvardžiuotinių formų sakytinėje kalboje vartojimo polinkių analizė, todėl taikomas analitinis aprašomasis bei statistinis metodas. Pastebėta, kad į įvardžiuotines formas sakytinėje kalboje kreipiamas mažas kalbininkų dėmesys, šia tema trūksta platesnių tyrimų ar teorinių metmenų. Šis darbas ir yra noras paskatinti tyrinėti ir didžiuotis lietuvių kalbos įvardžiuotinėmis formomis, taip pat parodyti jų svarbą dabartinėje sakytinėje lietuvių kalboje. Atlikus tyrimą, išryškėjo šios tendencijos: įvardžiuotinės formos sakytinės kalbos... [toliau žr. visą tekstą] / Language is the main indicator of human interaction. Radio and television are the most popular public media, therefore it is very important that the broadcast language would be lively, grammatically correct, that it would strengthen the existing language norms rather than brake them. The aim of this Master thesis is to research the forms and functions of pronominal forms used in spoken language by various groups of language users. The hypothesis formulated in this work is the following: in spoken language pronominal forms are used very rarely or no longer used; most often they are required only to construct complex definitions and names. This hypothesis is discussed in relation to the analysis of selected television and radio programs and in relation to the formulated conclusions. Research material includes nineteen programs, seven of which are radio programs (Pozicija, Prie pietų stalo, Ryto garsai, Žmonės ir idėjos, Dienos tema, Druska) and twelve television programs (Įžvalgos, Ką manai, Kultūros namai, Spaudos klubas, Požiūris, Nepatogūs pokalbiai). Full analysis of habits of use of pronominal forms in spoken language is presented in this thesis, thus analytical-descriptive and statistical methods are applied. It has been noted, that linguists pay little attention to pronominal forms in spoken language, and that there is a lack of wider research and theoretical outline of this field. This thesis also draws an intention to encourage the research and explore... [to full text]

Many positive outcomes have been documented for children with hearing loss utilizing current treatment approaches such as early identification and intervention, including appropriately fit sensory devices and communication modes that focus on listening and spoken language. However, challenges related to social-emotional development have been widely observed. The development of communication skills in children with hearing loss is impacted by many factors, including the degree of hearing loss, the child's age at onset and identification, the presence of other disabilities, and when the child receives intervention. While there are a variety of therapeutic options available for children with hearing loss to develop communication skills, listening and spoken language is of particular interest to parents with normal hearing. In addition to affecting social competence and participation, problems with social-emotional development are linked to poor academic performance. This study examined the social-emotional development of a small group of young children who communicated using listening and spoken language as measured by parent and caregiver report. Three psychosocial scales were used to evaluate the children's social-emotional development in comparison to peers. These results were analyzed within the context of other demographic variables. One of the five children was identified as facing problems with social-emotional development.