Nous avons observé depuis plusieurs années l’émergence des approches statistiques pour la traduction automatique. Cependant, l’efficacité des modèles construits est soumise aux variabilités inhérentes au langage naturel. Des études ont montré la présence de vocabulaires spécifique et général composant les corpus de textes de domaines spécialisés. Cette particularité peut être prise en charge par des ressources terminologiques comme les lexiques bilingues.Toutefois, nous pensons que si le vocabulaire est différent entre des textes spécialisés ou génériques, le contenu sémantique et la structure syntaxique peuvent aussi varier. Dans nos travaux,nous considérons la tâche d’adaptation aux domaines spécialisés pour la traduction automatique statistique selon deux axes majeurs : l’acquisition de lexiques bilingues et l’édition a posteriori de traductions issues de systèmes automatiques. Nous évaluons l’efficacité des approches proposées dans un contexte spécialisé : le domaine médical. Nos résultats sont comparés aux travaux précédents concernant cette tâche. De manière générale, la qualité des traductions issues de systèmes automatiques pour le domaine médical est améliorée par nos propositions. Des évaluations en oracle tendent à montrer qu’il existe une marge de progression importante / These last years have seen the development of statistical approaches for machine translation. Nevertheless, the intrinsic variations of the natural language act upon the quality of statistical models. Studies have shown that in-domain corpora containwords that can occur in out-of-domain corpora (common words), but also contain domain specific words. This particularity can be handled by terminological resources like bilingual lexicons. However, if the vocabulary differs between out and in-domain data, the syntactic and semantic content may also vary. In our work, we consider the task of domain adaptation for statistical machine translation through two majoraxes : bilingual lexicon acquisition and post-edition of machine translation outputs.We evaluate our approaches on the medical domain. The quality of automatic translations in the medical domain are improved and the results are compared to other works in this field. Oracle evaluations tend to show that further gains are still possible

[pt] O problema central desta tese está no letramento de alunos surdos, o que levou à reflexão sobre a constituição do pensamento através de signos e seus possíveis significados construídos socialmente. Faz-se a tentativa de preencher a lacuna apontada por Lebedeff (2010) sobre quais seriam as práticas pedagógicas advindas da necessidade discursiva da experiência visual da surdez e a quais eventos de letramento visual se referem esses discursos. Os objetivos específicos da pesquisa foram os seguintes: 1. Identificar as características das práticas pedagógicas dos instrutores surdos em atuação nas escolas-piloto de educação bilíngue no município do Rio de Janeiro; 2. Analisar as inter-relações entre as práticas pedagógicas dos instrutores surdos com o letramento visual ou o uso de recursos apoiados em imagens. 3. Caracterizar as práticas pedagógicas negociadas com o professor (ouvinte) junto ao instrutor surdo. Contudo, verificamos que há outras práticas que estão ocorrendo diante de fatores, tais como a forte tendência autobiográfica do surdo, manifesta na construção de narrativas surdas em uma linguagem própria, o que envolve não somente a língua de sinais ou língua escrita (portuguesa). Para a compreensão das negociações de sentido, parte-se para a linguagem de interações que é fator fundamental à compreensão da comunicação e da sociabilidade; estas não se concretizam somente por meios e suportes, mas também mediada por interpretações, anseios e conflitos. Diante dessa perspectiva, consideramos que os surdos encontram-se em um processo de desequilíbrio ritual e abordamos os jogos de interação em uma linguagem goffmaniana (1981, 1985, 1988, 2011). Motivou-nos abordar esse panorama complexo em escolas regulares onde foram validadas etapas metodológicas sucessivas: 1. Conhecimento do contexto arquitetônico, humano e tecnológico das escolas; 2. Apresentação da educação bilíngue ao diretor e coordenador pedagógico; 3. Acompanhamento das salas de recursos, pela pesquisa, de modo a documentar a bidocência por meio dos registros fílmicos; 4. Construção de material de formação continuada a partir da edição de fotografias e filmagens para espaço da formação com os Instrutores Surdos, Intérpretes de LIBRAS e Professores das Salas de Recursos; 5. Encontros anuais de Educação Bilíngue organizados para 400 a 600 participantes. A pesquisa envolveu também algumas providências quanto a: equipar as salas de recursos com materiais pertinentes à surdez; capacitar professores, instrutores surdos e intérpretes de LIBRAS; oferecer cursos de formação continuada em LIBRAS. Delinearam-se três estilos de aula: a. Expositiva em língua de sinais com o uso mais tradicional de apresentação de assunto ou tema com incremento de imagens; b. Informatizada em língua de sinais com o uso de outros meios como história em quadrinhos, desenhos, jogos e representações; c. Dramatizada ou teatralizada em língua de sinais com o uso de recursos cênicos. Considerando que a perspectiva fundamental da didática assume o processo ensino-aprendizagem em sua multidimensionalidade, ou seja, os aspectos político, humano e técnico, como nos indica Candau (1983, 2012a), enxergá-la junto a uma parte dos instrutores surdos de língua de sinais fez parte da proposta desta pesquisa-ação que demandou 69 incursões a campo com filmagens desses estilos de aula e de recursos usados em 12 escolas-piloto de Educação Bilíngue no Município do Rio de Janeiro. Justifica-se o esforço de análise e, principalmente, de catalogação dos artefatos encontrados nas escolas públicas municipais cariocas, cerne desta tese, na qual o pano de fundo não deixam de ser as narrativas surdas, porém não nos prendendo a elas, mas à arte dos surdos e de alguns ouvintes em criar suportes, recursos, práticas pedagógicas que criam outros modos de ensino-aprendizagem em prol de uma didática da invenção surda. / [en] The central concern in this thesis is in the literacy of deaf students, which led to the consideration of the constitution of thought through signs and its possible socially constructed meanings. It is focused on the attempt to fill the gap pointed by Lebedeff (2010) on which would the pedagogical practices be from the discursive need of the visual experience of deafness and to which visual literacy events these discourses refer to. The specific objectives of the research were: 1. To identify the characteristics of the pedagogical practices of the deaf instructors acting in pilot schools of bilingual education in the city of Rio de Janeiro; 2. To analyze the interrelationships between the pedagogical practices of the deaf instructors with visual literacy or the use of resources supported by images. 3. To describe the pedagogical practices negotiated with the teacher (listener) together with the deaf instructor. However, we found out that are other practices that are occurring in the face of these factors such as the strong autobiographical tendency of the deaf individuals, expressed by the construction of deaf narratives in their own language, which involves not only sign language or written language (Brazilian Portuguese). To understand the negotiations of meaning, we set out to the language of interactions, which is a key factor in understanding the communication and sociability; these are realized not only by means and media, but also mediated by interpretations, desires and conflict. Given this perspective we considered that the deaf individuals are in a process of ritual imbalance and we approached games of interaction in a Goffman language (1981, 1985, 1988, 2011). We were motivated to address this complex panorama, in research, in regular schools where successive methodological steps were validated: 1 Knowledge of architectural, human, and technological context of the schools; 2 Presentation of bilingual education for the director and pedagogical coordinator; 3 Monitoring of resource rooms, in order the research, to document two teachers (deaf and hearing) through the records on film (movie); 4 Construction of continuing education material from the photo editing and filming for training with the deaf instructors, sign language interpreters and teachers of resources rooms; 5. Bilingual Education Annual meetings arranged for 400 to 600 participants. The research also involved some steps as: equip the rooms with resource materials relevant to deafness; continuing training for teachers, deaf instructors and deaf sign language interpreters; offer continuing education courses in sign language. Outlining three styles of curricula: a. Lecture in sign language with the more traditional use of presentation of subject or theme with incremental images; b. Computerized lesson in sign language with the use of other media like comics, cartoons, games and representations; c. Dramatized lesson in sign language with the use of scenic resources. Whereas basic didactic perspective assumes teaching-learning process in its multidimensionality, i.e. the political, human and technical aspects, as indicated by Candau (1983, 2012a), it was part of the purpose of this action-research to comprehend it along with some of the Deaf Instructors of the sign language: the lesson styles and the resources used by 12 Deaf Instructors. They are those who managed to promote sign language and deaf culture, with greater effectiveness, as the largest investment of resources in pilot schools of Bilingual Education in the Municipality of Rio de Janeiro. The effort of analysis is justified and especially the cataloging of the artifacts found in the public municipal schools of Rio de Janeiro - core of this thesis - in which the background is nonetheless deaf narratives, but not holding on to them, but to the art of deaf individuals and some listeners in creating means, resources, pedagogical practices that create other approaches to teaching and learning towards a teaching of the deaf invention.

Cette étude est une approche des divers problèmes que soulève la traduction du figement. Elle se fond sur le cas concret de la traduction français-arabe en analysant un corpus bilingue constitué d’expressions figées en français (langue source) et de leur traduction en arabe (langue cible). Notre travail s’organise en deux parties : dans la première, nous étudions la notion de figement, les différentes structures qu’elle recouvre et la notion de traduction avec les concepts qui lui sont liés à travers les théories linguistiques et traductologiques du 20e siècle. Dans la deuxième partie, nous analysons notre corpus, d’abord pour vérifier qu’il représente bien la catégorie du figement, ensuite pour connaître les stratégies adoptées par le traducteur pour traduire les expressions figées. Le corpus est extrait à partir des articles français du Monde diplomatique et leur traduction en langue arabe dans le même journal / This study is an approach to various problems raised bu the translations of frozen expressions. It is founded on the specific case of the french-arabic translation by analyzing a bilingual corpus composed of idioms in French (source language) and their Arabic translation (targuet language). Our work is organized in two parts ; in the first we study the notion of rigidification and the various structures that it covers, and olso the notion of translation with related concepts through 20th century linguistic and translational theories. In the second part we analyze our corpus : first to make sure it really represents frozen expressions, then to know the strategies adopted by the translator to translate idioms. The corpus is extracted from the articles appeared in the French edition of the le Monde diplomatique and their translations into Arabic in the same newspaper

Les décisions sur les langues, ou politiques linguistiques familiales, constituent le cadre de cette recherche portant sur la transmission de l'anglais dans des familles francophones résidant en France. L'objectif est de mieux comprendre pourquoi et comment ces décisions sont prises par les parents, à travers l'étude de leurs représentations et de la façon dont ils rendent compte de leurs décisions. Les résultats montrent que les représentations des parents sur l'enseignement des langues sont négatives et qu'elles expliquent pourquoi ils sélectionnent un modèle d'enseignement bilingue pour leur enfant. Leurs représentations du bi/plurilinguisme et de l'anglais sont positives : ils souhaitent avant tout transmettre une langue qu'ils affectionnent et qui participe à leur identité en tant qu'individus plurilingues plus qu'ils ne la considèrent utile. Le vécu linguistique des parents, leur désir de bilinguisme pour l'enfant, les questionnements liés à ce qu'implique le choix d'une autre langue que sa langue native, l'investissement ainsi que la place centrale de l'enfant dans ces décisions permettent de mieux comprendre leurs politiques linguistiques. / Decisions regarding languages, or what is called family language policy, is the main focus of the present study which deals with the transmission of English in francophone families living in France. To better understand how their decisions are taken, this research aims at analyzing the underlying parents' representations and the way they account for their choices. The results show how negative their representations of foreign language teaching in France are, which explains why they tend to favor a bilingual education for their children. Their representations of bi/plurilingualism, along with the English language, are positive and point to the fact they want to pass on a language that they are particularly fond of and which is part of their identity as plurilingual individuals. Their experience with languages, the project of bilingualism for their children, the questions raised by what it means to speak a language other than the native language, their investment and the place of the child within their decision making process enable us to better understand their family language policies.

A pesquisa investiga as concepções e teorias subjacentes às práticas didáticas propostas por professores em uma classe de 1º ano do Ensino Fundamental em uma escola bilíngüe. O trabalho parte de uma desconstrução do mito de monolingüismo no Brasil, apontando para a pluralidade lingüística e cultural que constitui a sociedade brasileira apesar das políticas de planificação lingüística. Define os conceitos de bilingüismo, educação bilíngüe e escola bilíngüe apoiando-se nas definições multidimensionais propostas por Hamers, Blanc, Mackey, Valdés e Figueroa, que levam em conta não apenas a proficiência nas duas línguas, mas o uso das línguas em situações de comunicação, e o bilingüismo visto como um processo em construção ao invés de um produto acabado. Os contextos bilíngües presentes no Brasil são enumerados em escolas bilíngües indígenas, escolas LIBRAS-português para surdos, escolas de fronteiras nos países do MERCOSUL, escolas internacionais e escolas bilíngües de prestígio, apresentando alguns aspectos históricos e sociais relativos à presença de cada um desses contextos na sociedade, sobretudo a conscientização dos direitos de terceira geração, como os direitos lingüísticos. Adota uma metodologia de pesquisa etnográfica, analisando em nível micro-sociológico um contexto específico, uma sala de aula de 1º ano do Ensino Fundamental em uma escola bilíngüe de prestígio, levantando dados por meio de observações e entrevistas. Analisa as práticas didáticas propostas por professores encontradas no trabalho de campo e as teorias a elas subjacentes, identificando um programa de imersão baseado na teoria de aquisição natural das línguas proposta por Krashen (natural approach), que propõe uma aquisição inconsciente da língua, semelhante à língua materna, valorizando os aspectos afetivos, lúdicos e comunicativos da língua. Encontra uma ambigüidade nas práticas de alfabetização expressa por uma visão mais ampla de alfabetização em língua portuguesa do que na língua inglesa, e relaciona-a a aspectos culturais presentes na assimetria entre metodologia de ensino e na concepção de material didático em cada língua. Relaciona a realidade encontrada em sala de aula com aspectos macro-sociais em uma perspectiva de mútua influência entre escola e sociedade, observando que o aumento do interesse pelo ensino-aprendizagem de línguas hegemônicas pode tanto constituir uma forma de aprofundamento das desigualdades sociais quanto instrumentalizar os indivíduos para terem acesso a uma amplitude maior de conhecimentos historicamente construídos. / This research investigates the conceptions and theories underlying didactic practices proposed by teachers in a first year Elementary Education class at a bilingual school. This work stems from a deconstruction of the monolinguism myth in Brazil, pointing to the linguistic and cultural plurality which constitutes Brazilian society in spite of the policies on linguistic standardization. Defines the concepts of bilingualism, bilingual education and bilingual school, basing itself on the multidimensional definitions proposed by Hamers, Blanc, Mackey, Valdés and Figueroa, which take into consideration not only proficiency in the two languages, but the use of these languages in situations of communication, and bilingualism seen as a work-in-progress and not a finished product. The bilingual contexts present in Brazil are enumerated in indigenous bilingual schools, LIBRAS-Portuguese schools for the deaf, schools on the borders of MERCOSUL countries, international schools and bilingual schools of prestige, presenting some historical and social aspects relative to the presence of each of these contexts in society, above all the awareness of the rights of the third generation, such as linguistic rights. The research adopts an ethnographic research methodology, analyzing at a micro-sociological level a specific context a classroom of a first year Elementary School Education group at a prestigious bilingual school, gathering data through observations and interviews. Analyzes the didactic practices proposed by teachers found in the field of work and the underlying theories; identifying a program of immersion based on the theory of the natural acquisition of languages proposed by Krashen (natural approach), which considers an unconscious acquisition of language, similar to mother language, valuing the affective, playful and communicative aspects of the language. Finds an ambiguity in the practice of the literacy process expressed by a more ample view of literacy in the Portuguese language than in the English language that might be regarded to diverse cultural features present in the methodology of education and the conception of didactic material in each language. Relates the reality found in the classroom with macro-social aspects, in a perspective of mutual influence between the school and society, observing that the increase in interest for the learning of hegemonic languages can constitute a form of deepening social inequality, as well as provide individuals with tools to have access to a greater amplitude of knowledge historically constructed.

La traduction assistée par ordinateur a connu un grand essor dans les années 1990 avec l'arrivée des environnements à mémoires de traduction. Ces systèmes exploitent la redondance des textes techniques produits et traduits dans l'industrie, en proposant aux traducteurs de réutiliser des traductions archivées et d'accroître ainsi leur productivité. Ces mémoires utilisent, sans analyse, des segments de textes (le plus souvent des phrases entières) dont le découpage et l'alignement sont garantis par le traducteur.<br /><br />Pourtant ces mémoires recèlent des gisements d'information importants au niveau sous-phrastique dont les utilisateurs ne peuvent pas bénéficier. Le formalisme TransTree permet de représenter des correspondances sous-segmentales enchassées bilingues. Ces correspondances complexes (les amphigrammes) forment une structure arborescente exprimable en XML. Une transformation de surface conduit à une visualisation dynamique mettant en évidence les différents niveaux de correspondance entre sous-segments.<br /><br />TransTree s'accompagne d'une méthode générale de construction par voie statistique, fondée sur les arbres binaires de sécabilité. Cette méthode permet d'établir des amphigrammes à partir des correspondances entre mots typographiques. Il est possible d'abstraire des patrons de traduction (amphigrammes génériques) par classification des exemples rencontrés dans le corpus.<br /><br />Quelques expérimentations ont été effectuées pour valider le pouvoir d'expression du formalisme, explorer différentes options de construction et esquisser un algorithme de reconstitution d'un segment cible à partir d'un segment source inconnu avec la connaissance extraite des mémoires de traduction.

Thèse de doctorat : Celtique : Rennes 2 : 2006. / Bibliogr. f. 471-473. Annexes.

Les lexiques bilingues sont des ressources particulièrement utiles pour la Traduction Automatique et la Recherche d'Information Translingue. Leur construction manuelle nécessite une expertise forte dans les deux langues concernées et est un processus coûteux. Plusieurs méthodes automatiques ont été proposées comme une alternative, mais elles qui ne sont disponibles que dans un nombre limité de langues et leurs performances sont encore loin derrière la qualité des traductions manuelles.Notre travail porte sur l'extraction de ces lexiques bilingues à partir de corpus de textes parallèles et comparables, c'est à dire la reconnaissance et l'alignement d'un vocabulaire commun multilingue présent dans ces corpus.

La plupart des travaux en acquisition de lexiques bilingues à partir de corpus comparables reposent sur l'hypothèse distributionnelle qui a été étendue au scénario bilingue. Deux mots ont de fortes chances d'être en relation de traduction s'ils apparaissent dans les mêmes contextes lexicaux. Ce postulat suppose donc une définition claire et rigoureuse du contexte et une connaissance parfaite des indices contextuels. Or, la complexité et les spécificités de chaque langue font qu'il n'est pas aisé d'énoncer une telle définition qui garantisse une extraction de couples de traductions, efficace dans tous les cas de figure. Toute la difficulté réside dans la manière de définir, d'extraire et de comparer ces contextes dans le but de construire des lexiques bilingues fiables. Nous nous efforcerons tout au long des différents chapitres de cette thèse à essayer de mieux comprendre cette notion de contexte, pour ensuite l'étendre et l'adapter afin d'améliorer la qualité des lexiques bilingues. Une première partie des contributions vise à améliorer l'approche directe qui fait office de référence dans la communauté. Nous proposerons plusieurs manières d'aborder le contexte des mots pour mieux les caractériser. Dans la deuxième partie des contributions, nous commencerons par présenter une approche qui vise à améliorer l'approche par similarité inter-langue. Ensuite, une méthode nommée Q-Align, directement inspirée des systèmes de question/réponse sera présentée. Enfin, nous présenterons plusieurs transformations mathématiques et donc plusieurs représentations vectorielles, pour nous concentrer essentiellement sur celles que nous aurons choisi pour développer une nouvelle méthode d'alignement.

No description available.