O presente estudo analisa as contribuições do uso da música e de pragmatemas na retenção lexical por parte de três turmas de inglês como língua adicional em uma escola estadual na cidade de São João de Meriti (RJ). Para tal, constituiu três condições de pesquisa das quais fizeram parte doze alunos na turma 1 (T1); doze alunos na turma 2 (T2); e nove alunos na turma 3 (T3). A turma 1 (T1) participou de práticas pedagógicas com foco na música e sem foco explícito nos pragmatemas presentes na letra; a turma 2 (T2) de práticas pedagógicas com foco na música e nos pragmatemas da letra; e a turma 3 (T3) de práticas pedagógicas voltadas somente para o ensino-aprendizagem dos pragmatemas presentes na letra, sem o suporte da música. Anteriormente à intervenção, foi aplicado um questionário sobre os hábitos sócio-culturais dos participantes visando conhecer suas preferências musicais e perfil de letramento. Para analisar a relação entre o uso da música e de pragmatemas na retenção lexical, foram aplicados um pré-teste antes da intervenção e dois pós-testes com intervalo de 1 mês entre eles. Além disso, o estudo contou com uma entrevista semi-estruturada com os participantes, visando entender a percepção dos mesmos sobre as práticas utilizadas na intervenção. Os achados apontam que houve uma retenção ligeiramente superior nas condições T2 (música e foco nos pragmatemas) e T3 (foco exclusivo nos pragmatemas), com ligeira superioridade para a condição 3. O pragmatema recuperado com maior frequência foi "What's up", fato que pode ser parcialmente explicado pelo fato de ser pronunciado como uma palavra só e pelo número de letras que o representa ortograficamente. Estudos futuros poderão esclarecer se há significância estatística entre as diferenças encontradas assim como melhor explorar como o ensino explícito de unidades lexicais complexas pode contribuir para a retenção desses itens lexicais e consequentemente para o ensino-aprendizagem de línguas adicionais. A intervenção permitiu que os aprendizes passassem a entender que as palavras não ocorrem de modo isolado, mas que sempre caminham junto a outras. Permitiu também que a rotina pedagógica contemplasse a coconstrução do conhecimento, levando os aprendizes a reagir positivamente às práticas utilizadas no ensino de inglês, conforme relatos durante as entrevistas semi-estruturadas. Isso per se já justifica a utilização de práticas semelhantes e ilustra a contribuição do presente estudo / The present study analyzes the contributions of the use of music and complex lexical units to lexical retention in three classes of English as an additional language in a public school in the city of São João de Meriti (RJ). In order to do so, the research constituted three conditions in which twelve students participated of class 1 (T1), twelve students participated of class 2 (T2), and nine students participated of class 3 (T3). Class 1 (T1) participated in pedagogical practices with a focus on music and without explicit focus on institutionalized expressions present in the lyrics of the song, class 2 (T2) participated in classes with a focus on music and present in the lyrics of the selected song, and class 3 (T3) participated in pedagogical practices geared only to the teaching-learning of institutionalized expressions present in the lyrics of the song, without the support of music. Prior to the intervention, a questionnaire to uncover the socio-cultural habits of the participants was applied to the participants in order to know their preferences as well as their background, their literacy profile and their musical preference. To analyze the relationship between the use of music and institutionalized expressions in the participants lexical retention, a pretest was applied before the intervention and two post-tests were applied with an interval of one month between them. Furthermore, the study relied on semi-structured interviews with participants in order to understand their perceptions of the intervention. The findings indicate that there was a slightly higher retention in conditions T2 (music and focus on institutionalized expressions) and T3 (exclusive focus on institutionalized expressions), with a slight superiority for condition 3. The most frequently recovered institutionalized expression was "what's up", which can be partially explained by the fact that it is pronounced as one word and by the number of letters which are present in the expression. Future studies may clarify whether there is a statistical significance between the differences found in the conditions as well as how the explicit teaching of complex lexical units can contribute to the retention of lexical items and thus for the teaching and learning of additional. Furthermore, the intervention allowed learners to better understand that words do not occur in isolation, but also that they always occur in combination with others. Learners only responded positively to the practices used in the study, but also reported enjoying them during the semi structured interviews. That in itself seems enough to justify the use of similar practices and illustrates the contribution of this study

Cette recherche s’intéresse au langage préfabriqué (LP) dans le discours oral spontané d’apprenants de français langue étrangère (FLE). Le LP, étudié en linguistique sous l’angle du figement et de la conventionalité, intéresse aussi les psycholinguistes. Omniprésent dans le discours des locuteurs natifs, il sous-entend que la maîtrise d’une langue ne passe pas seulement par l’application des règles grammaticales et la création des phrases mais également par la mémorisation et l’automatisation des bouts de langue préconstruits. Notre objectif est double : étudier l’évolution de la connaissance du LP chez des apprenants du FLE et, le rôle que le LP joue dans leur fluidité verbale. Dix-sept apprenants ont réalisé à six mois d’intervalle deux tâches narratives orales, et leurs productions ont été enregistrées et transcrites deux fois. Six locuteurs natifs ont été soumis aux mêmes tâches une fois. Le LP a été identifié selon des critères linguistiques et psycholinguistiques. La fluence des productions a été évaluée au moyen de quatre mesures. Elle a été mise en relation avec l’emploi effectif des séquences préfabriquées afin de vérifier en quel sens ces dernières améliorent le discours oral des apprenants. Ainsi, il existe des différences individuelles très marquées entres les sujets testés et il est difficile de tirer des conclusions d’ordre général. Cependant, les apprenants de niveau débutant progressent dans l’emploi qu’ils font du LP, en quantité et en qualité. Ils peuvent s’appuyer sur les SP pour produire un discours plus fluide. Elles améliorent le segment de parole (longueur et complexification) et servent de stratégies (compensation, structuration et planification). / This thesis focuses on formulaic language in French as a Second Language (FSL) learners’ spontaneous spoken narratives. Formulaic language is made of multi-word units which are often perceived and used as simple lexical units. ESL studies have shown that even at advanced levels the command of formulaic language can be poor. The purpose of this thesis is twofold. First we intend to examine the evolution of formulaic language among FSL learners, and then the importance of formulaic language in learners’ fluency. Seventeen learners divided into three levels were asked over a period of six months to produce two oral narratives based on a cartoon and a silent film. Their narratives were recorded and transcribed as time 1 and time 2. The formulaic language which was found in the oral productions was identified according to linguistic and psycholinguistic criteria : phonological coherence, syntactic complexity, semantic opacity, frequence and systematicity of idiosyncratic errors and deviant items. The fluency was evaluated by means of four “classical” measurements : speech rate, length of run, rate of articulation, phonation time. Our analyses show that there are individual differences among subjects and that it is difficult to draw comprehensive conclusions. Yet, the results show a steady improvement in formulaic language among beginners, both quantitatively and qualitatively. Indeed, learners can rely on formulaic sequences to gain fluency and this in different manners. Formulaic sequences improve the speech run by making it longer and more complex. They can be used as strategies of compensation and planification of speech.

Training a Support Vector Machine (SVM) requires the solution of a very large quadratic programming (QP) optimization problem. Sequential Minimal Optimization (SMO) is a decomposition-based algorithm which breaks this large QP problem into a series of smallest possible QP problems. However, it still costs O(n2) computation time. In our SVM implementation, we can do training with huge data sets in a distributed manner (by breaking the dataset into chunks, then using Message Passing Interface (MPI) to distribute each chunk to a different machine and processing SVM training within each chunk). In addition, we moved the kernel calculation part in SVM classification to a graphics processing unit (GPU) which has zero scheduling overhead to create concurrent threads. In this thesis, we will take advantage of this GPU architecture to improve the classification performance of SVM.