There has been no commonsense on how to identify problem domain concerns in architectural modeling of software systems. Even, there is no commonly accepted method for modeling the Non-Functional Requirements (NFRs) effectively associated with the architectural aspects in the solution domain. This thesis introduces the use of a Machine Learning (ML) method based on Support Vector Machines to relate NFRs to classified &quot / architectural concerns&quot / in an automated way. This method uses Natural Language Processing techniques to fragment the plain NFR texts under the supervision of domain experts. The contribution of this approach lies in continuously applying ML techniques against previously discovered &ldquo / NFR - architectural concerns&rdquo / associations to improve the intelligence of repositories for requirements engineering. The study illustrates a charted roadmap and demonstrates the automated requirements engineering toolset for this roadmap. It also validates the approach and effectiveness of the toolset on the snapshot of a real-life project.

Being one of the most epoch making invention of the 21st century, computers have fundamentally altered every aspect of people&rsquo / s lives, including the education domain. Thus, for decades, computer-assisted language learning has received considerable amount of attention among researchers and language teachers by serving interactive, multi-sensory, and autonomous learning opportunities. In this respect, innumerable large or small-scale projects blazed a trail for other teachers to follow. The present study, therefore, is primarily concerned with the students&rsquo / attitudes towards computer-assisted language learning. Its main purpose is to investigate what the students&rsquo / attitudes are towards computer- assisted language learning (CALL) by also taking their attitude towards computer assisted learning (CAL) and foreign language learning (FLL) into consideration. Finally, factors affecting students&rsquo / attitudes and the relationships among computer assisted learning, computer assisted language learning and foreign language learning are also explored within the scope of the study. The findings demonstrate that most of the students have positive attitudes towards computer assisted learning, computer assisted language learning and foreign language learning. Age, grade, gender, years of studying English and prior CALL experience affect students&rsquo / attitudes. Moreover, students attitudes towards computer assisted language learning, computer- assisted language learning, and foreign language learning are, indeed, interrelated.

Thesis (Ph. D.)--University of Washington, 2001. / Vita. Includes bibliographical references (leaves 203-212).

Thesis (M.A.)--Miami University, Dept. of Spanish and Portguese, 2004. / Title from first page of PDF document. Includes bibliographical references (p. 46-52).

Thesis (PhD)--Macquarie University, Division of Linguistics & Psychology, Department of Linguistics, 2007. / Bibliography: 269-288.

Este trabalho busca analisar as representações construídas no discurso de alunos do Ensino Médio em uma escola pública. O objetivo é o de verificar os diferentes efeitos de sentido produzidos por discursos além do pedagógico e investigar sua eventual interferência no processo de identificação, tendo em vista as condições de produção do discurso da Escola Pública. Adotou-se uma abordagem transdisciplinar, à luz da Análise de Discurso (AD), Psicanálise, Filosofia e de Estudos Culturais, por meio do levantamento de alguns conceitos da AD e da conceituação de noções tais como as de identidade, representação e identificação. Partiu-se da hipótese central de que o sujeito, interpelado por diferentes discursos, a partir de um processo de identificação com relação ao falante de inglês, toma o idioma como um instrumento que lhe possibilitará ocupar o lugar desse falante. As condições de produção dos discursos trabalhados foram apontadas através de breves históricos da instrução pública no Brasil, da instituição da Língua Inglesa no currículo escolar e da sua posterior solidificação entre nós por meio de uma aliança cultural entre Brasil e Estados Unidos da América. O corpus foi formado por entrevistas gravadas em áudio, as quais foram transcritas e, posteriormente, analisadas. As análises enfocam as representações dos sujeitos acerca da instrução pública no Brasil, da aprendizagem de inglês na Escola Pública, da aprendizagem de inglês em Institutos de Idiomas, da cultura como facilitadora da aprendizagem e da língua inglesa como um bem comercializável. Dentre os discursos que constituem os sujeitos, destacamos, nessas análises, o discurso oficial da escola pública, o da mídia, o da cultura e o político-econômico. Dessas análises depreende-se que os alunos, interpelados por esses e outros discursos, são submetidos a identificações com formações que os fazem ocupar posições sujeito diversas, até por vezes contraditórias. Sugere-se, ademais, que o sujeito aprendiz de inglês no contexto de uma escola pública identifica-se com a representação do lugar ocupado pelo falante de inglês. Observa-se, no entanto, que, apesar de expressar vontade de aprender inglês, nem sempre sua possível identificação permite a ressignificação do sujeito em um processo de aprendizagem de um idioma estrangeiro. / English LaThis paper analyzes the representations built in the discourse of high-schoolers from a public school in Brazil. The aim is to recognize the effects of meaning possibly being produced by discourses other than the pedagogic and to investigate whether they have a role in the identification process, given the conditions of production of the discourse of the Public School. A transdisciplinary approach is adopted, including concepts from Discourse Analysis, Psychoanalysis, Philosophy and Cultural Studies. We thus bring in some concepts of Discourse Analysis and also conceptualize notions such as identity, representation and identification. The central hypothesis is that the subject, interpellated by different discourses, from an identification process related to the English Speaker, considers language as an instrument that will enable him/her to take the place that an English speaker has. The conditions of production of the various discourses studied in this paper are presented via brief historical surveys of the public education in Brazil, of the inclusion of the English language in the Brazilian curriculum and of the circumstances that caused English to be considered the most important foreign language within the learning system after an agreement between cultural agencies of Brazil and the United States of America. The corpus consists of audio taped interviews that were transcribed and analyzed. The analysis focuses the representations, by the subjects, of the public education in Brazil, of learning English in Public School, of learning English in Language Institutes, of culture as a facilitator of learning and of English as a commodity. Among discourses that constitute the subjects, we highlight the official discourse of public school, the discourse of media, the discourse of culture and the political-economic discourse. These analyses show that students, interpellated by these and other discourses, identify themselves with formations that make them occupy different, even contradictory, subject positions. We suggest that the subject, an English learner whose context is that of a Brazilian public school, identifies him/herself with his/her representation of the place that an English speaker takes. It is to be noted, however, that notwithstanding the subjects expression of his/her willingness to learn English, not always his/her (possible) identification allows for re-signifying within a process of learning a foreign language.

Esta tese dedica-se a observar, a partir de um teste experimental, o processamento (momento em que os aprendizes manipulam os dados da língua estrangeira) da forma uno/a do espanhol, utilizada para referências ora mais ora menos genéricas, bem como o processo (sucessão de estados ou de mudanças, no nosso caso, ao longo da aplicação dos testes) pelo qual passam esses aprendizes de língua espanhola - de diferentes grupos, com diferentes características, que têm como língua materna o português brasileiro - ao enfrentarem a tarefa solicitada. O aprendiz, ao passar por diferentes momentos de processamento da forma uno/a, que vemos como um processo, às vezes consegue e outras vezes não consegue compreender os sentidos que essas formas podem recobrir no espanhol, o que ocorre tanto com aquele aprendiz que recebeu instrução formal sobre a forma uno/a como com aquele que não a recebeu. Com base nos dados aos quais o aprendiz se viu exposto, este por vezes produz e por vezes não produz enunciados adequados na língua estrangeira, assim como também produz novas possibilidades compatíveis com esse fenômeno tão peculiar que é a interlíngua, que concebemos como variável e sincrética. Com base na experiência realizada, defendemos nesta tese, que produção e compreensão são processos sistêmicos e simultâneos realizados pelos aprendizes em cada processamento, valendo-se de seus conhecimentos prévios de natureza gramatical e pragmático-discursiva, tanto da língua estrangeira que está aprendendo, quanto da sua língua materna, bem como de suas experiências anteriores e de seus conhecimentos de mundo. Isto significa dizer que o aprendiz, tal como qualquer sujeito, possui \"(...) um cérebro para a linguagem, não só para a gramática\" (Possenti, 1992: 75) e, por isso, passamos a considerá-lo um sujeito complexo. O modelo interpretativo que defendemos tem sua base epistemológica na teoria da complexidade, visto que, no momento do processamento, as informações que vêm da gramática, do discurso, da pragmática e de outros conhecimentos prévios interagem simultaneamente como um sistema dinâmico, porque obedecem a uma organização que alterna entre estabilização e desestabilização, imprevisibilidade, instabilidade e descontinuidade / This thesis presents experimental research regarding the processing (moment in which beginners manipulate the foreign language data) and the process (sequence of conditions or changes; in our study, a sequence of tests) of the use of the Spanish form uno/a (used for more or less generic references) by different groups of Brazilian students learning Spanish as a foreign language in Brazil. Different outcomes result when the beginner undergoes different processing experiences of uno/a. Both the beginner who received the formal instruction about uno/a and the one who did not sometimes do and sometimes do not understand the different meanings of these forms in Spanish. In spite of the input to which the beginners are exposed to, they may or not produce adequate utterances in the foreign language and they can also create new uses consistent with the interlanguage phenomenon, conceived as variable and syncretic. In this thesis, production and comprehension are defended as systemic and simultaneous processes performed by the beginners in each processing procedure, articulated with their previous grammatical, discursive and pragmatical knowledge both of the foreign and native language, including also their previous experiences and knowledge of world. That is to say that the beginner, as anyone else, possess \"(...) a brain for the language, not only for the grammar\" (Possenti, 1992: 75) and, therefore, can be viewed as a complex being. The interpretative model defended in this work has its epistemological basis in the complexity theory. Second language processing is seen as a complex dynamic system which simultaneously relate grammar, discourse, pragmatics and other previous knowledge into an organization that changes itself from stability to instability, imprevisibility and discontinuity

Die vorliegende Arbeit beschreibt das Design und die Implementierung eines Algorithmus zur Flexion. Es wird am Beispiel des Deutschen eine konkrete Implementierung entwickelt. Hierfür findet zunächst eine ausführliche Analyse der Flexion des Deutschen statt, bevor ein Verfahren erarbeitet wird, das sprachunabhängig ist und somit prinzipiell auf andere Sprachen übertragen werden kann. Die tatsächliche Machbarkeit des Verfahrens wird anhand von Beispielen nachgewiesen. Die hohe Komplexität der Aufgabe führt allerdings dazu, dass es in der Praxis zu Abstrichen bei der Qualität der flektierten Wortformen kommt. Dies ist insbesondere deswegen der Fall, da das entwickelte System auch ihm unbekannte Grundformen flektiert.

This work explores multilingual speech synthesis. We compare three models based on Tacotron that utilize various levels of parameter sharing. Two of them follow recent multilingual text-to-speech systems. The first one makes use of a fully-shared encoder and an adversarial classifier that removes speaker-dependent information from the encoder. The other uses language-specific encoders. We introduce a new approach that combines the best of both previous methods. It enables effective parameter sharing using a meta- learning technique, preserves encoder's flexibility, and actively removes speaker-specific information in the encoder. We compare the three models on two tasks. The first one aims at joint multilingual training on ten languages and reveals their knowledge-sharing abilities. The second concerns code-switching. We show that our model effectively shares information across languages, and according to a subjective evaluation test, it produces more natural and accurate code-switching speech.

Current dialogue systems typically consist of separate components, which are manu- ally engineered to a large part and need extensive annotation. End-to-end trainable sys- tems exist but produce lower-quality, unreliable outputs. The recent transformer-based pre-trained language models such as GPT-2 brought considerable progress to language modelling, but they rely on huge amounts of textual data, which are not available for common dialogue domains. Therefore, training these models runs a high risk of overfit- ting. To overcome these obstacles, we propose a novel end-to-end dialogue system called AuGPT. We add auxiliary training objectives to use training data more efficiently, and we use massive data augmentation via back-translation and pretraining on multiple datasets to increase data volume and diversity. We evaluate our system using automatic methods (corpus-based metrics, user simulation), human evaluation as part of the DSTC 9 shared task challenge (where our system placed 3rd out of 10), as well as extensive manual error analysis. Our method substantially outperforms the baseline on the MultiWOZ bench- mark and shows competitive results with state-of-the-art end-to-end dialogue systems. 1