The general aim of this thesis was to test the effects of paralinguistic (emotional) and prior contextual (topical) cues on perception of poorly specified visual, auditory, and audiovisual speech. The specific purposes were to (1) examine if facially displayed emotions can facilitate speechreading performance; (2) to study the mechanism for such facilitation; (3) to map information-processing factors that are involved in processing of poorly specified speech; and (4) to present a comprehensive conceptual framework for speech perception, with specification of the signal being considered. Experi¬mental and correlational designs were used, and 399 normal-hearing adults participated in seven experiments. The main conclusions are summarised as follows. (a) Speechreading can be facilitated by paralinguistic information as constituted by facial displayed emotions. (b) The facilitatory effect of emitted emotional cues is mediated by their degree of specification in transmission and ambiguity as percepts; and by how distinct the perceived emotions combined with topical cues are as cues for lexical access. (c) The facially displayed emotions affect speech perception by conveying semantic cues; no effect via enhanced articulatory distinctiveness, nor of emotion-related state in the perceiver is needed for facilitation. (d) The combined findings suggest that emotional and topical cues provide constraints for activation spreading in the lexicon. (e) Both bottom-up and top-down factors are associated with perception of poorly specified speech, indicating that variation in information-processing abilities is a crucial factor for perception if there is paucity in sensory input. A conceptual framework for speech perception, comprising specification of the linguistic and paralinguistic information, as well as distinctiveness of primes, is presented. Generalisations of the findings to other forms of paralanguage and language processing are discussed.

Cette thèse présente un environnement ouvert et souple pour l'acquisition automatique d'expressions multimots (MWE) à partir de corpus textuels monolingues. Cette recherche est motivée par l'importance des MWE pour les applications du TALN. Après avoir brièvement présenté les modules de l'environnement, le mémoire présente des résultats d'évaluation intrinsèque en utilisant deux applications: la lexicographie assistée par ordinateur et la traduction automatique statistique. Ces deux applications peuvent bénéficier de l'acquisition automatique de MWE, et les expressions acquises automatiquement à partir de corpus peuvent à la fois les accélérer et améliorer leur qualité. Les résultats prometteurs de nos expériences nous encouragent à mener des recherches ultérieures sur la façon optimale d'intégrer le traitement des MWE dans ces applications et dans bien d'autres

Faces provide cues for judgments regarding the emotional state of individuals. Using signal-detection methodology and a standardized stimulus set, the overall aim of the present dissertation was to investigate the detection of emotional facial expressions (i.e., angry and happy faces) with neutral expressions as the nontarget stimuli. Study I showed a happy-superiority effect and a bias towards reporting happiness in female faces. As work progressed, questions arose regarding whether the emotional stimuli were equal with regard to perceived strength of emotion, and whether the neutral faces were perceived as neutral. To further investigate the effect of stimulus quality on the obtained findings, Study II was designed such that the facial stimuli were rated on scales of happy-sad, angry-friendly, and emotionality. Results showed that ‘neutral’ facial expressions were not rated as neutral, and that there was a greater perceived distance between happy and neutral faces than between angry and neutral faces. These results were used to adjust the detectability measures to compensate for the varying distances of the angry and happy stimuli from the neutral stimuli in the emotional space. The happy-superiority effect was weakened, while an angry-female disadvantage remained. However, as these results were based upon different participant groups for detection and emotional rating, Study III was designed to investigate whether the results from Studies I and II could be replicated in a design where the same participants performed both tasks. Again, the results showed the non-neutrality of ‘neutral’ expressions and that happiness was more easily detected than anger, as shown in general emotion as well as specific emotion detection. Taken together, the overall results of the present dissertation demonstrate a happy-superiority effect that was greater for female than male faces, that angry-female faces were the most difficult to detect, and a bias to report female faces as happy. / At the time of the doctoral defense, the following papers were unpublished and had a status as follows: Paper 1: In press. Paper 2: Manuscript. Paper 3: Manuscript.

En este estudio, mostraremos principalmente la importancia que las imágenes formadas por los componentes fraseológicos pueden adquirir desde un punto de vista no sólo puramente lingüístico, sino también psicolingüístico si se contempla el posible carácter analizable y transparente de muchas EI. Explicaremos cómo se puede utilizar esta iconicidad para fomentar el desarrollo de estrategias de aprendizaje (especialmente vinculadas a la descodificación metafórica o a la asociación icónica con expresiones de la LM) que suponen para el alumno un alto grado de implicación cognitiva. Teniendo en cuenta dos factores importantes para la memorización de nuevas EI -su grado de transparencia semántica y su grado de conexión icónica con la LM-, propondremos en un apartado experimental investigar acerca del impacto memorístico e interlingüístico que se puede atribuir al trabajo de enfoque icónico y verificar la influencia que la combinación de los dos factores citados puede ejercer en este ámbito. / This study will mainly show how important the images formed by the phraseological components could be from not only a merely linguistic point of view but also from a psycholinguistic one, if we consider the possible analyzable and transparent nature of most idioms. We will explain how to use this iconic aspect of idioms in order to encourage the development of learning strategies (especially related to the metaphorical decodification or to the iconic association with L1 idioms) that involve a high degree of cognitive implication for the L2 learner. Considering two important factors for the memorization of L2 idioms - its degree of semantic transparency and its degree of iconic connection with the L1 -, we propose, in an experimental part, to investigate the memory and interlinguistic effects of an image-oriented work and to verify the influence that the combination of the two factors mentioned before could have in this field.

L'ouvrage unifie les paradigmes du darwinisme universel, de la psycholinguistique développementale et de la linguistique computationnelle afin de fournir un récit nouveau d'ontogénie de structures linguistiques chez les agents humains ou bien artificiels. La thèse est précédée d'un volume supplémentaire intitulé «Fondements conceptuels» qui présente une théorie de l'évolution intra-mentale qui pose l'hypothèse que l'ontogénie d'un esprit individuel peut être interprétée et même simulée comme un processus impliquant la réplication, la variation et la sélection des structures qui encodent de l'information. La dissertation elle-même présente quatre simulations distinctes abordant quatre problèmes distincts. La 0ème simulation illustre comment l'optimisation évolutionnist pourrait conduire à la découverte d'idées utiles concernant l'énigme cryptologique connue sous le nom de Voynich Manuscript. La première simulation montre comment la théorie des prototypes, les architectures symboliques vectorielles et l'optimisation évolutionniste peuvent être mutuellement combinées afin de produire une nouvelle méthode d'apprentissage automatique supervisée. La deuxième simulation utilise une approche similaire pour démontrer que l'optimisation évolutive peut découvrir des constellations minimalistes et légères de marqueurs de partie de discours. La dernière simulation vise le «saint graal» de la linguistique computationnelle, c'est-à-dire le problème de «l'induction grammaticale» et montre que le problème peut être potentiellement résolu en utilisant une stratégie évolutive capable de combler l'écart entre le domaine subsymbolique et le domaine symbolique. / Our thesis unifes paradigms of universal darwinism, developmental psycholinguistics and computational linguistics in order to furnish a novel account of language development in human as well as artificial language-acquiring agents. Thesis is preceded by a supplementary volume called "Conceptual Foundations" which presents a so-called "Theory of Intramental Evolution" which postulates that ontogeny of an individual mind can be interpreted and even simulated as a process involving replication, variation and selection of information-encoding cognitive structures. The dissertation itself presents four distinct simulations addressing four distinct problems. Zeroth simulation illustrates how evolutionary optimization could lead to discovery of useful insights concerning the cryptological riddle known as Voynich Manuscript. The first simulation shows how theory of prototypes, vector symbolic architectures and evolutionary optimization can be mutually combined in order to yield a novel supervised machine-learning method. Second simulation uses a similiar approach in order to indicate that evolutionary optimization can discover minimalist and lightweight constellations of part-of-speech taggers. The last simulation targets the "holy grail" of computational linguistics, i.e. the problem of "grammar induction" and shows that the problem can be potentially solved by using an evolutionary strategy able to bridge the gap between subsymbolic realm of vector spaces and symbolic realm of grammar-representing regular expressions.

Orientador: Alexandre César Rodrigues da Silva / Resumo: Com a disseminação de dispositivos eletrônicos cada vez menores e o advento de novas tecnologias. A busca por métodos de minimização de funções booleanas tem sido a base para eletrônica digital. Neste trabalho apresenta-se a implementação da primeira fase do método Quine-McCluskey Estendido que utiliza-se de estruturas AND-XOR-XNOR para a geração de implicantes primos. O objetivo do trabalho foi comprovar que, na maioria das vezes, a implementação de uma função Booleana utilizando expressões AND-XOR-XNOR requerem menor quantidade de termos produtos, quando comparado com implementação com expressões AND-OR. A fase de cobertura dos mintermos em ambos os métodos foi formulada como um problema de programação linear inteira 0 e 1 que através do programa Lp_solve obteve a solução de menor custo. Na comparação da eficiência dos métodos foram analisados os custos dos circuitos mínimos gerados, a quantidade de memória utilizada e o tempo de execução. Com os resultados obtidos pode-se concluir que, para a maioria dos casos executados, o método Quine-McCluskey Estendido gera uma solução de menor custo. No entanto, com relação ao desempenho computacional (tempo de execução e memória), o método Quine-McCluskey Estendido apresentou-se inferior se comparado ao Quine-McCluskey. / Mestre

This thesis investigates the central issues underlying graph analysis, namely, scalability and quality. We first study the incremental problems for graph queries, which aim to compute the changes to the old query answer, in response to the updates to the input graph. The incremental problem is called bounded if its cost is decided by the sizes of the query and the changes only. No matter how desirable, however, our first results are negative: for common graph queries such as graph traversal, connectivity, keyword search and pattern matching, their incremental problems are unbounded. In light of the negative results, we propose two new characterizations for the effectiveness of incremental computation, and show that the incremental computations above can still be effectively conducted, by either reducing the computations on big graphs to small data, or incrementalizing batch algorithms by minimizing unnecessary recomputation. We next study the problems with regards to improving the quality of the graphs. To uniquely identify entities represented by vertices in a graph, we propose a class of keys that are recursively defined in terms of graph patterns, and are interpreted with subgraph isomorphism. As an application, we study the entity matching problem, which is to find all pairs of entities in a graph that are identified by a given set of keys. Although the problem is proved to be intractable, and cannot be parallelized in logarithmic rounds, we provide two parallel scalable algorithms for it. In addition, to catch numeric inconsistencies in real-life graphs, we extend graph functional dependencies with linear arithmetic expressions and comparison predicates, referred to as NGDs. Indeed, NGDs strike a balance between expressivity and complexity, since if we allow non-linear arithmetic expressions, even of degree at most 2, the satisfiability and implication problems become undecidable. A localizable incremental algorithm is developed to detect errors using NGDs, where the cost is determined by small neighbors of nodes in the updates instead of the entire graph. Finally, a rule-based method to clean graphs is proposed. We extend graph entity dependencies (GEDs) as data quality rules. Given a graph, a set of GEDs and a block of ground truth, we fix violations of GEDs in the graph by combining data repairing and object identification. The method finds certain fixes to errors detected by GEDs, i.e., as long as the GEDs and the ground truth are correct, the fixes are assured correct as their logical consequences. Several fundamental results underlying the method are established, and an algorithm is developed to implement the method. We also parallelize the method and guarantee to reduce its running time with the increase of processors.

[pt] Este trabalho investiga os estereótipos sobre o brasileiro construídos pelos alunos estrangeiros do curso de Português como segunda língua (PL2E) da PUC-Rio. Objetivou-se identificar as expressões qualificativas, representadas em especial pelos adjetivos, verbos ou advérbios, e também substantivos, em relação à cultura brasileira, analisar essas avaliações dos alunos antes e depois de estarem imersos no Brasil e mostrar como elas podem interferir na comunicação intercultural. A análise revelou que: (i) grande parte dos alunos estrangeiros já tinham construído uma imagem prévia e estereotipada do brasileiro e do Brasil; (ii) os estereótipos construídos têm relação com as informações colhidas pelos alunos nos meios de massa como televisão, jornais, internet e também através de amigos brasileiros ou não e parentes; (iii) as imagens construídas pelos estrangeiros são positivas em relação ao brasileiro, e negativas quanto ao Brasil; (iv) a atitude positiva diante da imagem dos brasileiros e a exposição à cultura brasileira facilitam a interação nas situações interculturais desse aluno com brasileiros. / [en] This paper investigates the stereotypes about the Brazilian people built by the foreign students in the Portuguese as a Second Language Course at PUC-Rio. The objective was to identify the qualificative expressions in relation to the Brazilian culture - specially represented by the adjectives, verbs and adverbs -, analyse these evaluations from the students before and after having been immersed in Brazil, and show how they can interfere in the intercultural communication. The analysis revealed that: (i) a big part of the foreign students had built a stereotyped image of the Brazilians and Brazil; (ii) the stereotypes have to do with the information collected by the students in the media, i.e. TV, newspaper, internet and also through Brazilian friends, Brazilians in general and relatives; (iii) the images built by the foreigners are positive in relation to the Brazilian and negative when it comes to Brazil; (iv) the positive attitude in relation to the image of the Brazilians and the exposure to the Brazilian culture make the foreigners interaction in the intercultural situations with Brazilian students easy.

O alinhamento léxico automático é uma tarefa essencial para as técnicas de tradução de máquina empíricas modernas. A abordagem gerativa não-supervisionado têm sido substituída recentemente por uma abordagem discriminativa supervisionada que facilite inclusão de conhecimento linguístico de uma diversidade de fontes. Dentro deste contexto, este trabalho descreve uma série alinhadores léxicos discriminativos que incorporam heurísticas de pós-processamento com o objetivo de melhorar o desempenho dos mesmos para expressões multi-palavra, que constituem um dos desafios da área de processamento de linguagens naturais atualmente. A avaliação é realizada utilizando um gold-standard obtido a partir da anotação de um corpus paralelo de legendas de filmes. Os alinhadores propostos apresentam um desempenho superior tanto ao obtido por uma baseline quanto ao obtido por um alinhador gerativo do estado-da-arte (Giza++), tanto no caso geral quanto para as expressões foco do trabalho. / Lexical alignment is an essential task for modern empirical machine translation techniques. The unsupervised generative approach is being replaced by a supervised, discriminative one that considerably facilitates the inclusion of linguistic knowledge from several sources. Given this context, the present work describes a series of discriminative lexical aligners that incorporate post-processing heuristics with the goal of improving the quality of the alignments of multiword expressions, which is one of the major challanges in natural language processing today. The evaluation is conducted using a gold-standard obtained from a movie subtitle parallel corpus. The aligners proposed show an alignment quality that is superior both to our baseline and to a state-of-the-art generative aligner (Giza++), for the general case as well as for the expressions that are the focus of this work.

O presente trabalho trata basicamente do conteúdo expressões algébricas no ensino fundamental. Motivados pela crença de que a ausência do método de argumentação matemática e do pensamento genérico no atual ensino de Matemática têm se apresentado muito nociva ao estudo de expressões algébricas, desenvolvemos, em uma turma de sétima série, uma proposta didática onde apresentamos uma sequência de atividades que enfatizam estes dois aspectos. Tais atividades servem de pré-requisitos para o estudo de expressões algébricas. Inclui-se neste trabalho uma análise dos PCN para o ensino fundamental e uma análise crítica de livros didáticos de sétima série/oitavo ano sobre expressões algébricas. A partir destas análises, percebemos, nos livros didáticos, o inadequado emprego (no nosso ponto de vista) de polinômios neste nível de ensino, o que resultou na elaboração de um capítulo de conteúdo matemático versando sobre polinômios, dedicada a professores de Matemática da Educação Básica. / This text is closely related to the subject algebraic expressions in Fundamental School. Believing that the absence of the mathematical method and of what we call generical thought in the treatment of mathematics in the first school years causes much harm in the student’s mathematical education, mainly when one deals with algebraic expressions, we applied in a 7th grade class and present here a sequence of activities which give emphasis to those two aspects mentioned above and which prepare the students for the study of algebraic expressions. Additionally we include here an analysis of part of the PCN (Brazilian Curriculum Recommendations) and of nine 7th grade-school books, namely, the part related to algebraic expressions. Finally, we include a chapter about polynomials which was written for school teachers, motivated by the inadequate use (for this level of mathematical education, at least in our point of view) of polynomials in the school books which we have analyzed.