No segmento criminal, uma ferramenta digital capaz de comparar informações sobre suspeitos criminais, restringindo a lista de potenciais criminosos, representa um grande avanço tecnológico. O desafio é conseguir recuperar imagens similares em banco de dados a partir do retrato falado de um suspeito, sendo assim possível localizar as imagens de criminosos no banco de faces da polícia. A importância desse desafio deve-se ao fato que na maioria dos casos, o retrato falado gerado, a partir da descrição verbal de testemunhas é o ponto de partida para o desfecho de muitas investigações. Em vista disto, este trabalho tem por objetivo apresentar um sistema computacional capaz de buscar imagens de faces humanas de um banco de faces através dos esboços das imagens de face, a partir de um retrato falado. Foram feitos retratos falados para comparar com os esboços das imagens das faces humanas presentes em banco de dados e foram recuperadas as mais semelhantes. Na metodologia proposta, as imagens de face passam por um processamento chamado Misturograma, que resulta no esboço da imagem de face, que é utilizado para se comparar com os retratos falados obtidos. O algoritmo proposto usa distância Euclidiana para comparação de semelhança e wavelets de Haar como descritor das imagens. Foram utilizadas as métricas estatísticas de avaliação de desempenho: Recall X Precision e Cumulative Match Score. A metodologia proposta obteve uma precisão de 95% de acertos nas buscas realizadas em Cross Validation, dessa forma mostra-se um método viável para recuperação de imagens a partir de retratos falados. / In the criminal segment, a digital tool able to compare information on criminal suspects, restricting the list of potential criminals, represents a major technological breakthrough. The challenge is to retrieve similar images in the database and thus from the sketch of a suspect you can find images of criminals in the police faces database. The importance of this challenge is due to the fact that in most cases, the sketch generated from the verbal description of witnesses is the starting point for the outcome of many investigations. In view of this, this paper aims to develop a computer system able to search for images of human faces from a database of faces using face images of outlines from a sketch. Sketches were made to compare with the outlines of the images of human faces present in the database and the most similar were retrieved. In the proposed methodology, the face images go through a process called Misturograma, resulting in the outlines of the face image, which is used to compare the obtained sketches. The proposed algorithm uses Euclidean distance to compare the similarity and Haar wavelets as a descriptor of the images. We used the statistical metrics for performance evaluation: Recall X Precision and Cumulative Match Score. The proposed method achieved an accuracy of 95% in carried out searches in Cross Validation thus proves to be a viable method for recovering images from sketches.

Contemporary studies of media Arabic often pass over issues of media form and the broader relevance of language use. The present thesis addresses these issues directly by examining the language used in Jordanian non-government radio programmes. It examines recordings and transcriptions of a range of programme genres - primarily, morning talk shows and 'service programmes' (barāmiž ḳadamātiyya), and Islamic advice programmes, both of which feature significant audience input via call-ins. The data are examined through an interpretive form of discourse analysis, drawing on linguistic anthropological theory that analyses language as a form of performance, through comparison of radio programmes as 'units of interaction'. This is supported by sociolinguistic data obtained from the recordings, including phoneme frequency analysis, in addition to the author's experience of 6 months of fieldwork in Jordan in 2014-15. The analysis focuses on four major themes: (1) the influence of media context, specifically the sonic exclusivity and temporal evanescence of radio, on language use, as well as the impact of digital media; (2) the indexicality of certain locally salient sociolinguistic variables, and the use to which they are put in radio talk; (3) the role of language in constructing the identity, or persona, of broadcasters; and (4) the role of language in constructing and validating authoritative discourse, in particular that of Islamic texts and scripture in religious programming. Through its analysis of these themes, using selected recording excerpts as demonstrative case studies, this thesis shows that specific strategies of Arabic use in the radio setting crucially affect both the publics - the addressed audiences - of radio talk, as well as the frameworks of participation in this talk - how and to what extent broadcasters and members of the public can participate in mediated discourse. The results demonstrate the unique value of an interpretive study of linguistic performance for highlighting broader social issues, including the inclusion and exclusion of particular segments of the society through linguistic strategies - Jordanians versus non-Jordanians, Ammanis versus non-Ammanis, and pious Muslims versus non-believers; and the use of language to reassert, or occasionally challenge, dominant ideologies and discourses, such as those of gender, nationalism, and religion. This study thus contributes an examination of contemporary Jordanian non-government radio language in its social and political context - something which has not been attempted before, and which provides important insights regarding both the nature of contemporary Arabic media language and its broader social and cultural import.

All current theories of spoken word recognition (e.g., Allopenna et al., 1998; McClelland & Elman, 1986; Norris, 1994) suggest that any part of a target word triggers activation of candidate words. Visual world paradigm studies have relied on the linking hypothesis that the probability of looking at the referent of a word directly tracks the word’s level of activation (e.g., Allopenna et al., 1998). However, how much information is needed to trigger a saccade to a visual representation of the word’s referent? To address this question, the present study manipulated the number and location of shared segments between the target and competitor words. Experimental evidence is provided by two visual world paradigm experiments on Japanese, using natural and synthesized speech. In both experiments, cohort competitor pictures were not fixated more than unrelated distractor pictures unless the cohort competitor shares the initial CVC with the target. Bayesian analyses provide strong support for the null hypothesis that shorter overlap does not affect eye movements. The results suggest that a listener needs to accumulate enough evidence for a word before a saccade is generated. The human data were validated by an interactive computational model (TRACE: McClelland & Elman, 1986). The model was adapted to Japanese language to examine whether the TRACE model predicts competitor effects that fit human data. The model predicted that there should be effects when words share any amount with a target which confirms the current theory. However, the model did not fit the human data unless there is longer overlap between words. This indicates that eye movements are not as closely tied to fixation probabilities of lexical representations as previously believed. The present study suggests that looking at a referent of a word is a decision, made when the word’s activation exceeds a context-specific threshold. Subthreshold activations do not drive saccades. The present study conclude that decision-making processes need to be incorporated in models linking word activation to eye movements.

This thesis project addresses the machine learning (ML) modelling aspects of the problem of automatically extracting typological linguistic information of natural languages spoken in South Asia from annotated descriptive grammars. Without getting stuck into the theory and methods of Natural Language Processing (NLP), the focus has been to develop and test a machine learning (ML) model dedicated to the information extraction part. Starting with the existing state-of-the-art frameworks to get labelled training data through the structured representation of the descriptive grammars, the problem has been modelled as a supervised ML classification task where the annotated text is provided as input and the objective is to classify the input to one of the pre-learned labels. The approach has been to systematically explore the data to develop understanding of the problem domain and then evaluate a set of four potential ML algorithms using predetermined performance metrics namely: accuracy, recall, precision and f-score. It turned out that the problem splits up into two independent classification tasks: binary classification task and multiclass classification task. The four selected algorithms: Decision Trees, Naïve Bayes, Support VectorMachines, and Logistic Regression belonging to both linear and non-linear families ofML models are independently trained and compared for both classification tasks. Using stratified 10-fold cross validation performance metrics are measured and the candidate algorithms  are compared. Logistic Regression provided overall best results with DecisionTree as the close follow up. Finally, the Logistic Regression model was selected for further fine tuning and used in a web demo for typological information extraction tool developed to show the usability of the ML model in the field.

Ce travail s’est donné comme but d’analyser un corpus de presse écrite constitué de sept titres allemands et de sept titres français datant de 2006, en vue de répertorier tous les phénomènes de langue de conception parlée [terminologie Koch-Oesterreicher 1985 et 1990, aussi: oralité] à l’intérieur de ce corpus et ce afin d’observer le degré de perméabilité des différents titres vis-à-vis de ces phénomènes et les normes scripturales journalistiques en vigueur dans les deux pays. Pour ce faire, il a été dressé un inventaire de catégories permettant de passer au peigne fin le corpus pour en recueillir, dans une banque de données, les phrases-tokens présentant au moins un trait de langue de CONCEPTION parlée- d‘oralité. Ces catégories servant de filtre ont été arrêtées suite à l‘élaboration, dans la première partie du travail, d‘une liste contenant les caractéristiques observables dans les langue! s orales, de conception parlée, des deux pays. Les résultats sont interprétés dans une optique tridimensionnelle: celle du cadre énonciatif, celle du contrat de communication et celle de l‘oralité fictive. Une analyse plus détaillée de chaque catégorie se trouve par ailleurs en annexes [annexes n° 1, tome II]. Il s‘avère alors que les magazines pour jeunes sont très réceptifs en matière de phénomènes d‘oralité et que la langue de la presse allemande y est plus ouverte que son correspondant français / This paper has aimed to analyze a corpus of print media made of seven German and seven French 2006 newspapers and magazines in order to make an inventory of all those phenomena of spoken language [terminology by Koch-Oesterreicher 1985 and 1990, also: oral language] inside it so as to look at how pervious the different titles are vis-à-vis these phenomena and so as to judge about the journalistic norms currently ruling in both countries. To do so, the author has elaborated an inventory of different categories allowing to comb the corpus to gather those sentences [tokens] into a data base which presented at least one item of spoken language. The aforesaid categories working as filters have been acquired as a result of the elaboration, in the first part of this paper, of a list of caracteristics observable in the spoken languages of the two countries

Cette recherche propose une analyse multimodale du positionnement interactionnel ou stance-taking. Le corpus de travail, filmé, transcrit et annoté par nos soins dans trois logiciels compatibles (CLAN, PRAAT, ELAN), est une collection de discussions semi-guidées sur le thème de l’environnement (2h 20 min). Les 16 locuteurs sont des étudiants locuteurs natifs d’anglais britannique qui discutent par deux et entre amis. Dans cette recherche, nous adoptons une définition large du « langage », en y incluant l’ensemble des ressources sémiotiques verbales et non-verbales mobilisées pour la co-construction dynamique et intersubjective du sens au cours de l’interaction orale. Nous montrons que les locuteurs intègrent une grande variété de ressources verbales (segments, énoncés), mais aussi vocales (intonation) et mimo-posturo-gestuelles (gestes, expressions du visage), en les synchronisant de manière tant simultanée que séquentielle, pour prendre position vis-à-vis de leur interlocuteur. Au plan théorique, notre approche multi-niveaux et multimodale tisse des liens entre théories françaises de l’énonciation (Benveniste, 1966, Morel et Danon-Boileau, 1998), théorie discursive-fonctionnelle du stance-taking (Kärkkäinen, 2006, Du Bois, 2007), analyse conversationnelle multimodale (C. Goodwin et M.H. Goodwin, 1992, Mondada, 2007), anthropologie linguistique (Ochs, 1996), et étude de la gestualité (Kendon, 2004, Müller, 2004, Streeck, 2009) ; au plan méthodologique, nous combinons analyse qualitative et codage systématique. Notre thèse pose d’abord les bases théoriques et méthodologiques d’une étude multimodale des stances (Partie 1), puis propose la possibilité d’un marquage visuel du positionnement intersubjectif (Partie 2), avant de montrer comment les locuteurs intègrent mots et syntaxe, voix, visage et corps pour prendre position en interaction (Partie 3). / In this research, we propose a multimodal analysis of stance-taking based a collection of semi-guided discussions between pairs of friends who discuss environmental issues (2h 20 min). All 16 speakers are university students who are native speakers of British English. We filmed, transcribed and annotated this video corpus in three compatible software tools, CLAN, PRAAT and ELAN. In this research, we defend a broad understanding of “language”, defined as encompassing all verbal and non-verbal semiotic resources involved in the dynamic and intersubjective co-construction of meaning during spoken interaction. We show that speakers integrate a wide range of verbal resources (segments, utterances) as well as vocal (intonation) and visual ones (gestures, postures and facial expressions), and synchronize these resources simultaneously and sequentially so as to take stances with respect to their interlocutors. On a theoretical level, our multi-level, multimodal approach brings together French utterer-centred approaches to language (Benveniste, 1966, Morel and Danon-Boileau, 1998), discursive-functional theories of stance-taking (Kärkkäinen, 2006, Du Bois, 2007), multimodal conversation analysis (C. Goodwin and M.H. Goodwin, 1992, Mondada, 2007), linguistic anthropology (Ochs, 1996) and gesture studies (Kendon, 2004, Müller, 2004, Streeck, 2009); our methodology combines qualitative analysis with systematic coding. This thesis starts with laying the theoretical and methodological bases for a multimodal study of stance-taking (Part 1); it then proposes that some gestures and facial expressions can be used as intersubjective visual stance markers (Part 2), before showing how speakers integrate words and syntax, voice, facial expressions, gestures and physical posture to take stance in interaction (Part 3).

Cette recherche vise à étudier et rendre-compte du fonctionnement des interactions orales entre pairs, en classe d’anglais L2, au niveau du collège en France. L’enjeu est donc, à partir d’une description détaillée de tâches (au sens didactique du terme) et de l’analyse des productions des apprenants, de documenter la notion de compétence d’interaction orale. L’approche retenue relève de la recherche-action et s’appuie sur des données empiriques recueillies dans des classes. Après avoir mené plusieurs études exploratoires sur des groupes d’apprenants, un protocole de recherche a été mis en place pour collecter des données longitudinales sur un échantillon 24 dyades, soit 48 élèves de 3ème. Les participants ont mené deux tâches d’interaction orale en début d’année puis en fin d’année scolaire, après avoir bénéficié, d’une formation spécifique à l’interaction orale. Leurs performances ont été filmées (18h) et transcrites sous CLAN pour être analysées. Au plan théorique, des liens sont tissés entre plusieurs cadres. Le premier est la didactique des langues qui retient la perspective actionnelle telle que décrite dans le CECRL (2001). Je m’appuie également sur la théorie socioculturelle, issue des travaux de Vygotsky (1978), puis de Lantolf (2000) qui avancent que l’interaction sociale ne se contente pas de faciliter l’apprentissage, mais est au cœur du processus de développement. Enfin les outils de l’analyse conversationnelle sent convoqués afin d’étudier la construction des échanges de manière systématique. Ce sont les questions de formation spécifique des apprenants à l’interaction orale, les résultats auxquels il est raisonnable de s’attendre dans ce contexte, et enfin les modalités d’évaluation de cette compétence en milieu institutionnel qui sont au cœur du sujet. / This research explores the potential of teenage peer spoken interactions in ESL classrooms in France. The focus here is on analysing task-based activities and learners’ performances, in order to improve our understanding of interactional competence. My approach is based on Action Research and a collection of empirical data from language classrooms. Several studies were conducted before setting up a protocol for a longitudinal corpus of video-recorded conversations between 24 dyads of peers. The 48 lower intermediate participants performed two tasks at the beginning and end of the school year. Specific training was offered in order to improve strategic spoken interaction. The 18 hours of verbal and non-verbal outcome was transcribed on CLAN and analysed according to the Conversational Analysis method. On a theoretical level, my approach brings together the European tradition of language didactics based on the CEFR (2001) and the sociocultural theory that takes its roots in Vygotsky’s (1978) work, followed by Lantolf (2000) who deals more specifically with SLA. According to these authors, social interaction not only promotes learning but is also at the core of a development process. This thesis examines the effects of training on communication strategies, the expectable outcomes and assessments of such a competence in an institutional environment.

This Statement of Purpose does not give a history of hip-hop or hip-hop poetry but rather how this particular poet fits into the current phase of hip-hop and performance poetry. In it, I discuss and explain the new pro-working class hip-hop performance poetic. This includes extensive discussion of how rap poetry conveys meaning through sound. Also discussed is the socioeconomic context for the suburban southwestern topics found in the manuscript. This statement is a parallel piece to the manuscript itself in that it explains a brown kid from San Bernardino’s journey of connecting words and music for purposes of personal expression and potentially the betterment of his community. This statement also discusses my aesthetic beliefs and preferences as a sound poet with a background in performance and music.

Orientador: Egisvanda Isys de Almeida Sandes / Banca: Elena Fernandez de Molina Ortes / Banca: Isabel Gretel María Eres Fernández / Banca: Kelly Cristiane Henschel Pobbe de Carvalho / Banca: Nildiceia Aparecida Rocha / Resumo: Este trabalho teve como objeto de estudo o ensino de pronúncia do espanhol como língua estrangeira (ELE). A discussão parte da observação de que o ensino de pronúncia nem sempre foi considerado relevante no ensino de línguas estrangeiras e de que durante muito tempo, os materiais didáticos abordaram a pronúncia por meio de exercícios e atividades descontextualizadas e de forma isolada em relação a outros aspectos da língua, como o léxico e a gramática. Para que ocorra uma comunicação fluída e inteligível, recomenda-se que o ensino da pronúncia se configure dentro de certa sistematicidade, mediante a definição de objetivos, conteúdos, atividades, exercícios e estratégias de correção específicas para cada nível, além de um trabalho que considere a relevância dos elementos suprassegmentais como ritmo, acento e entonação, para o estabelecimento de sentidos nos intercâmbios comunicativos. Nesse sentido, propõem-se as seguintes indagações: Como o ensino da pronúncia é abordado pelos diferentes métodos e abordagens de ensino de línguas estrangeiras? Quais conhecimentos são necessários para seu ensino e aprendizagem/aquisição, considerando o contexto da formação docente? Para tanto, o objetivo geral da pesquisa foi compreender como o ensino da pronúncia é retratado teoricamente pelos diferentes métodos e abordagens de ensino de línguas estrangeiras e pelos professores que o lecionam no ensino médio nas aulas de ELE. Assim sendo, especificamente, tem-se que: i) realizar uma revisão te... (Resumo completo, clicar acesso eletrônico abaixo) / Abstract: This work had as object of study the teaching of pronunciation of Spanish as a foreign language (SFL). The discussion is based on the observation that pronunciation teaching was not always consider relevant in foreign language teaching and that for a long time, the teaching materials addresses pronunciation through exercises and activities that were out of context and isolated from others aspects of the languages, such as lexicon and grammar. For a fluent and intelligible communication to take place, it is recommended that the teaching of pronunciation be configured with a certain system, by defining objectives, contents, activities, exercises and correction strategies specific to each level, in addition to a work that considers the relevance of suprasegmentally elements such as rhythm, accent and intonation, for the establishment of the sense in communicative exchanges. In this sense, the following questions are proposed: How is the teaching of pronunciation addressed by the different methods and approaches of the foreign language teachings? Which knowledge are needed for the teaching and learning /acquisition, taking in consideration the context of the teacher education? To this end, the general objectives of the research was to understand how pronunciation teaching is described theoretically by the different methods and approaches of the foreign language teachings and by the teachers who teach it in high school in the SFL classes. Thus, specifically, must to: i) carry out ... (Complete abstract click electronic access below) / Resumen: Este trabajo tuvo como objetivo de estudio la enseñanza de la pronunciación del español como lengua extranjera (ELE). La discusión parte de la observación de que la enseñanza de la pronunciación no siempre fue considerada relevante en la enseñanza de lenguas extranjeras y de que, durante mucho tiempo, los materiales didácticos abordaron la pronunciación por medio de ejercicios y actividades descontextualizadas y de forma aislada con relación a otros aspectos de la lengua, como el léxico y la gramática. Para que ocurra una comunicación fluida e inteligible, se recomienda que la enseñanza de la pronunciación se configure dentro de cierta sistematización, mediante la definición de objetivos, contenidos, actividades, ejercicios y estrategias de corrección específicas para cada nivel, además de un trabajo que considere la relevancia de los elementos suprasegmentales como ritmo, acento y entonación, para el estabelecimiento de sentidos en los intercambios comunicativos. De esta manera, se proponen las siguientes indagaciones: ¿Cómo la enseñanza de la pronunciación es abordada por los diferentes métodos y abordajes de enseñanza de lenguas extranjeras? ¿Cuáles conocimientos son necesarios para su enseñanza y aprendizaje/adquisición, considerando el contexto de la formación docente? Por tanto, el objetivo general de la investigación fue comprender como la enseñanza de la pronunciación es retratada teóricamente por los diferentes métodos y abordajes de enseñanza de lenguas extranjeras,... (Resumen completo clicar acceso eletrônico abajo) / Doutor

In 2003 the Ministry of Education, Culture, Sports, Science and Technology (MEXT) unveiled their new junior high school (JHS) English as a Foreign Language (EFL) policy, which focused strongly on oral communication. Although there is evidence of policy noncompliance in schools until now there has been no English language research on the attitudes or practices of Japanese teachers of English (JTEs), or the views of the students, and their parents in regards to teaching/learning English speaking skills. The research, based on JHSs in a mid-sized Japanese city (pop. 475,000 approx.), focused predominantly on JTEs, but also included students, and their parents. Focus group sessions, questionnaires, and one-on-one interviews were used to collect data. The study reveals that learning English speaking skills is considered important, but passing the senior high school (SHS) entrance examination is the main concern and so, test impact from the SHS entrance examination exerts the greatest pressure on JHS JTEs. The JTEs also perceive themselves as facing other issues such as student motivation, JTE speaking proficiency, and large class sizes. Another finding is that JTEs appear to receive inadequate training– pre- and inservice– resulting in issues, such as a reliance on traditional methods (yakudoku), which are not in accordance with MEXT’s intentions, and JTE proficiency test achievement levels lower than those desired by MEXT. As a result of these issues gaps exist between MEXT JHS EFL policies and actual teaching practices, and have unfortunately led to a situation where JTEs believe that MEXT does not care about or understand the teaching environment. The study concludes that implementation of MEXT’s policy require a better match between the SHS entrance examination and JHS EFL policy, a decrease in class sizes, and JTEs receiving more adequate training. A more positive relationship between MEXT and JTEs would result from these two groups working collaboratively when designing JHS EFL policies and could better achieve a match between the SHS entrance examination and JHS EFL policy.