Thesis (MA)--University of Stellenbosch, 2006. / Bibliography. Also available via the Internet.

We present a system to determine content similarity of documents. More specifically, our goal is to identify book chapters that are translations of the same original chapter; this task requires identification of not only the different topics in the documents but also the particular flow of these topics. We experiment with different representations employing n-grams of lexical chains and test these representations on a corpus of approximately 1000 chapters gathered from books with multiple parallel translations. Our representations include the cosine similarity of attribute vectors of n-grams of lexical chains, the cosine similarity of tf*idf-weighted keywords, and the cosine similarity of unweighted lexical chains (unigrams of lexical chains) as well as multiplicative combinations of the similarity measures produced by these approaches. Our results identify fourgrams of unordered lexical chains as a particularly useful representation for text similarity evaluation.

Orientador: Bento Carlos Dias da Silva / Banca: Clotilde de Almeida Azevedo Murakawa / Banca: Patrícia Tosqui Lucks / Resumo: A cidade de Ibitinga destaca-se nacionalmente no ramo dos bordados, com sua economia voltada quase que exclusivamente à produção dessas peças. Tendo esse contexto como pano de fundo, esta dissertação discute todo o processo linguístico e linguístico-computacional de construção de ontoléxicos - constructos formais cuja natureza léxico-conceitual possibilita o desenvolvimento de investigações teóricas (estudos lexicogramaticais) e aplicadas (construção de recursos lexicográficos e para o processamento computacional de informação textual disponível na Web). Em particular, constrói um ontoléxico exploratório que entrelaça conceitos e itens lexicais do domínio léxico-conceitual da Indústria do Bordado de Ibitinga. O embasamento teórico-metodológico assenta-se nos estudos de semântica lexical (wordnets) e de frames (framenets), pura e computacional, de processamento automático de língua natural e de ontologias. / Abstract: Ibitinga, which is a Brazilian town nationally known for its outstanding position in the embroidery business, has its economy almost totally based on the production of a broad selection of embroideries. With this embroidery industry in the backdrop, this master thesis describes the whole linguistic and computational-linguistic process of construction of ontolexicons - formal constructs whose lexical-conceptual nature allows for both theoretical (lexical-grammar construction) and applied (lexicographical and ontolexical resource construction) research. In particular, the study focuses on the design and implementation of a toy ontolexicon for the Ibitinga Embroydery Industry lexical-conceptual domain. Its theoretical foundations have drawn heavily on lexical semantics (wordnets), frame semantics (framenets), natural language processing, and ontologies. / Mestre

Étudier la liaison et les erreurs enfantines qu’elle entraîne (le nami, un zami) offre une perspective privilégiée sur le développement des représentations lexicales. Des travaux récents (Chevrot, et al., 2007 ; Dugua et al., 2009 ; Chevrot et al., à paraître) ont corroboré un modèle (Chevrot et al., 2009) rendant compte des étapes développementales jalonnant la segmentation des noms précédés d’une liaison. Nous avons mis en place, durant ce doctorat, trois démarches empiriques examinant des hypothèses issues de ce modèle.L’observation de 18094 groupes nominaux déterminant-nom issus d’un corpus de 118 heures de parole adressée à cinq enfants entre 1 et 3 ans fait apparaître trois tendances : 86% contiennent un nom à initiale consonantique (un garçon), la liaison (un ami) n’apparaît que dans 5% des GN, les consonnes enchainées sur l’initiale vocalique des noms sont principalement les liaisons /n/, /z/, /t/ ou le /l/ de l’élision (l’ami). La rareté des liaisons est compatible avec la complexité de leur acquisition. La prévalence des noms à initiale consonantique dans l’input est compatible avec la tendance enfantine à faire correspondre l’initiale du nom avec une consonne et ainsi à produire des segmentations de type /zami/, /nami/ ou /lami/ pour le mot ami. Cette étude aura permis de confirmer une hypothèse sur trois du modèle.Une expérience chronométrique menée chez 60 enfants de 5-6 ans montre que ces derniers ont mémorisé dans leur lexique aussi bien des variantes de type /nami/ et /zarbr/ pour les noms ami et arbre que des séquences fréquentes contenant ces variantes (un-ami, des-arbres). La même étude chez 36 adultes indique ceux-ci mémorisent dans leur lexique des séquences entières, sans retrouver de traces des variantes.Enfin, nous avons exploré l'hypothèse de Morin & Kaye (1982) qui suggéraient un statut particulier à la liaison /z/ : en effet, elle semble pouvoir être inscrite dans notre lexique comme un morphème du pluriel. De ce fait, elle bénéficierait d'un encodage encore plus abstrait, au niveau morphologique.Ces résultats sont discutés dans la perspective des théories basée sur l’usage (Tomasello, 2003) postulant un lexique constitué d’unités de longueurs et de niveaux d’abstraction variables. / Studying the liaison in French and childish mistakes it entails (un nami, un zami) offers a unique perspective on the formation of lexical representations. Recent work (Chevrot and al., 2007; Dugua and al, 2009;.. Chevrot and al, forthcoming) corroborated a model (Chevrot et al., 2009) reflecting the developmental stages during which segmentation of nouns that are preceded by a liaison. We have implemented during this PhD, three empirical approaches to examine the hypotheses which emerge from this model.Observing 18094 déterminant-noun groups from a corpus of 118 hours of speech addressed to five children between 1 and 3 years reveals three trends: 86% contain a noun with an initial consonant (un garçon), the liaison (un ami) appears only in 5% of NG's [for "Noun-Groups"], the consonants which are attached to the initial vowel of nouns are mainly the liaisons / n /, / z /, / t / or the elision / l / (l’ami). The scarcity of the liaison is consistent with the complexity of acquiring it. The prevalence of nouuns with an initial consonant in the input is compatible with children's tendency to match the initial of the name with a consonant and thus to produce segmentations like / zami /, / nami / or / lami / for the word ami. This study has confirmed one of the three hypothesis of the model.A chronometer experiment on 60 5-6 years old children shows that they have stored in their lexicon as well variants like / nami / and / zarbr / for nouns ami and arbre, as frequent sequences containing these variants (un-ami, des-arbres). The same study on 36 adults indicatesthat they store in their lexicon entire sequences, without finding any trace of these variants.Finally, we explored the hypothesis of Morin & Kaye (1982)which suggested that the liaison / z / had a special status : indeed, it seems to be written in our lexicon as a morpheme of the plural form. Therefore, it would benefit from an even more abstract encoding, morphologically.These results are discussed from the perspective of theories based on the common use (Tomasello, 2003) wich states that lexicon consists of units with varying lengths and levels of abstraction.

Non-native tones are considered challenging for adult second language speakers to perceive and produce. The current study examined the effect of a laboratory-based intensive training in improving American English speakers’ tone production. Participants’ task was to repeat Mandarin words after the model. There were two conditions in the experiment: in one condition, participants did not get any external feedback; whereas in the other condition, participants received detailed visual feedback, which was the pitch contour of their tone production alongside the native version. Eight participants completed training with no feedback and another eight participants were trained with visual feedback. Results revealed that participants in both groups did not improve their tone production after training, and participants trained with visual feedback did not show more improvement than those trained with no feedback. Given the lack of improvement in participants’ tone production after training, methodological and theoretical limitations with respect to the use of a repetition-based training paradigm are discussed.

Com base em alguns dos principais trabalhos existentes na literatura gerativista sobre a "alternância do dativo" do inglês, identificamos e sistematizamos neste trabalho as principais restrições que, segundo a literatura, atuam na construção dativa (construção dativa preposicionada ou a de objeto duplo), bem como algumas das principais propostas de análise. Nosso intuito é mostrar como a evolução do estudo de uma construção particular envolve, muitas vezes de modo indireto e não explícito, aspectos importantes, mais gerais, da teoria lingüística. No caso particular deste trabalho, procuramos mostrar como o estudo da alternância nos permite identificar, entre outras coisas: (a) concepções importantes da relação entre sintaxe e léxico, e (b) os tipos de argumentação que podem ser utilizados para distinguir processos sintáticos de processos lexicais, bem como processos lexicais de vários tipos entre si. Quanto ao percurso das análises sobre a alternância dativa e as principais concepções teóricas que foram discutidas, procuramos enfatizar o fato de que há a gradativa evolução da teoria ao longo dos anos. Tomamos como ponto de partida para nossa discussão sobre a alternância a tese de doutorado de Richard Oehrle (1976). Oehrle toma como ponto de partida para sua discussão a visão "puramente transformacional" da alternância oferecida pelo "modelo padrão" da teoria gerativa. Desta perspectiva, a alternância dativa é um processo "fortemente produtivo": processos transformacionais afetam constituintes sem ter acesso a propriedades léxico-semânticas particulares deles. No entanto, por meio de uma análise detalhada da construção, Oehrle foi o primeiro a demonstrar que tal visão estava equivocada: a alternância dativa é um processo "semiprodutivo", sujeita a diferentes restrições que parecem operar sobre subclasses lexicais de verbos, fazendo que itens muito semelhantes em significado ora possam, ora não, aparecer numa ou noutra das estruturas dativas do inglês. As demais abordagens analisadas neste trabalho concentram-se na tentativa de localizar adequadamente os fatores que regem a alternância, seja dentro do léxico, ou da sintaxe, ou na relação entre os dois. / On the basis of some of the main works in the generative literature about the English Dative Alternation, we try to identify and systematize the main constraints that act upon the dative constructions, as well as some of the main proposals of analysis. Our aim is to show how the evolution of the study of a particular construction may involve, sometimes in an indirect, inexplicit manner, important general aspects of the linguistic theory. Here we intend to show how the study of the dative alternation allows us to identify: a) important conceptions of the relation between syntax and lexical, and (b) the types of argumentation that can be used to distinguish syntactic processes from the lexical processes, and lexical processes of different kinds among themselves. As regards the succession of the analyses of the dative alternation, we try to emphasize the fact that there is a gradual evolution of the theory throughout the years. We begin our discussion with the doctorate thesis of Richard Oehrle (1976). Oehrle takes as a starting point a "purely transformational" analysis of the dative alternation, as might be provided by the "standard model" of the generative theory. From the perspective of such an analysis, the dative alternation should be a very productive process: transformational processes affect constituents without having access to their particular lexical-semantics properties. However, through a very detailed analysis of the alternation, Oehrle was the first to show that this view is inadequate: the dative alternation is a semi-productive process, subject to different constraints that seem to operate on lexical subclasses of verbs; in particular, even items that are very similar in meaning sometimes can, sometimes cannot, appear in the alternation. The other approaches that will be analyzed in this dissertation are all attempts to identify the right place where different aspects of the construction should be accounted for - either within the lexicon, or within syntax, or in the relation between both.

Este trabalho investiga a interação da Estrutura Informacional (EI) de duas construções marcadas do português - as clivadas e os tópicos contrastivos - com propriedades semânticas e pragmáticas associadas a estas duas estruturas. No que concerne às clivadas, o trabalho investiga de que modo a EI interage com quatro propriedades "de significado" comumente associadas a esta construção: (a) a leitura "especificacional" da sentença como um todo; (b) o "efeito de exaustividade" associado ao constituinte focalizado; (c) o caráter "pressuposicional" da oração clivada; e (d) o caráter "denegador" da sentença clivada como um todo dentro do discurso. A principal conclusão é de que tanto a propriedade (b) quanto a (d) não parecem ser "inerentes" à clivada; em particular, a propriedade (b) - que é analisada como um subproduto de um tipo particular de foco, o "foco identificacional", no influente trabalho de Kiss (1998) - não parece ser um aspecto convencionalizado da EI, mas sim um efeito da interação entre as propriedades (a) e (c) - estas sim, "convencionais" relativamente às clivadas. Quanto aos tópicos contrastivos, o trabalho concentra-se no impacto de sua EI na estrutura do discurso; mais especificamente, este estudo procura esclarecer qual o papel da EI dos tópicos contrastivos - e o papel da EI em geral - no estabelecimento de "relações retóricas" como a de "contraste", normalmente equacionada ao significado da conjunção adversativa mas. A conclusão é de que a EI dos tópicos contrastivos, em conjunto com a EI não-marcada, é que parece induzir a relações retóricas de "contraste" - independentemente de elementos externos à EI, como expressões que, à semelhança do mas, veiculam convencionalmente contraste. Deste modo, teorias que se valham da EI para estruturar o discurso - como a de Büring (2003), por exemplo - parecem ser mais adequadas para lidar com este tipo de fenômeno do que outras que depositam a maior parte do poder descritivo nas relações retóricas em si, como a de Asher & Lascarides (2003). / This work investigates the interaction of the Information Structure (IS) of two marked constructions of Portuguese - the clefts and the contrastive topics - between semantic and pragmatic properties associated with those constructions. Concerning clefts, this work investigates how IS interacts with four "properties of meaning" commonly associated with cleft sentences: (a) the "specificational" reading of the sentence as a whole; (b) the "exhaustivity effect" associated with the focalized constituent; (c) the "presuppositional" character of the cleft clause; and (d) the "denying" character of the cleft sentence as a whole in the discourse. The main conclusion is that neither property (b) nor property (d) seem to be "inherent" properties of clefts; in particular, property (b) - which is analyzed as a sub-product of a particular kind of focus, the "identificational focus", in Kiss' (1998) influential paper - doesn't seem to be a conventionalized aspect of IS, but instead a product of the interaction between properties (a) and (c) - which are actually "conventionalized" with respect to clefts. As for contrastive topics, this work concentrates on the impact of its IS in discourse structure; more specifically, this paper tries to clarify what is the role of the IS of contrastive topics - and of IS in general - in the establishment of "rhetorical relations" such as "contrast", usually identified with the meaning of conjunctions like but. The conclusion is that the IS of contrastive topics, together with the non-marked IS, seems to induce rhetorical relations like "contrast" - independently of elements external to IS, like expressions that, as but, induce contrast conventionally. Thus theories that use IS to structure the discourse - like Büring's (2003), for instance - seem to be more adequate to deal with this kind of phenomenon than theories that place much of the descriptive power in rhetorical relations itself, as in Asher & Lascarides (2003).

131 f. / Submitted by Cynthia Nascimento (cyngabe@ufba.br) on 2013-06-25T16:08:57Z No. of bitstreams: 2 Itatismara Valverde Medeiros - anexos.pdf: 24000062 bytes, checksum: 023f504f9897bd43db9a5000d5496389 (MD5) Itatismara Valverde Medeiros.pdf: 652534 bytes, checksum: ad4d24c75e228fe6766bfeace2442ce1 (MD5) / Approved for entry into archive by Alda Lima da Silva(sivalda@ufba.br) on 2013-06-27T20:59:15Z (GMT) No. of bitstreams: 2 Itatismara Valverde Medeiros - anexos.pdf: 24000062 bytes, checksum: 023f504f9897bd43db9a5000d5496389 (MD5) Itatismara Valverde Medeiros.pdf: 652534 bytes, checksum: ad4d24c75e228fe6766bfeace2442ce1 (MD5) / Made available in DSpace on 2013-06-27T20:59:15Z (GMT). No. of bitstreams: 2 Itatismara Valverde Medeiros - anexos.pdf: 24000062 bytes, checksum: 023f504f9897bd43db9a5000d5496389 (MD5) Itatismara Valverde Medeiros.pdf: 652534 bytes, checksum: ad4d24c75e228fe6766bfeace2442ce1 (MD5) / O Trabalho ora apresentado representa um estudo diacrônico estrutural do campo lexical da“sexualidade de religiosos” em cantigas satíricas medievais. Objetiva-se apresentar, descrever e analisar os subcampos do campo lexical da sexualidade de religiosos, textualizados na poesia lírica galego-portuguesa e, a partir daí, discutir os aspectos sociais e os valores morais do comportamento erótico-sexual de religiosos consagrados no léxico e textualizados nas referidas produções. Para tal descrição, adotou-se a teoria dos campos, tomando-se por base os ensinamentos dos campos lexicais apresentados pelos estudiosos Pottier, Greimas e, em especial, a teoria da Lexemática, de Coseriu. O corpus do Trabalho abrange 18 cantigas de escárnio e maldizer galego-portuguesas editadas, criticamente, por Rodrigues Lapa (1970) e por Graça Videira Lopes (2002). Para o desenvolvimento do Estudo, foram selecionadas e organizadas as lexias referentes à sexualidade de religiosos em subcampos e microcampos, seguidas de suas análises sêmicas. Tal teoria entende a língua como constituída por uma estrutura lexical que reflete a epistema com a qual uma determinada comunidade forja a sua realidade. Dessa forma, tem-se uma ideia mais aproximada das relações de conjunção e disjunção responsáveis por ligar e opor as unidades e os subcampos do campo, além de uma provável classificação conceitual que explique as diversas escalas ou níveis organizacionais das unidades. Este Estudo contribui, parcialmente, para ampliar o conhecimento do léxico da língua portuguesa medieval e para a apreensão de sua história, modelada pela dinâmica das comunidades linguísticas, em seus processos de socialização. Além do mais, constrói ferramentas válidas para o conhecimento e descrição, ainda que parcial, do funcionamento do campo lexical da sexualidade que se expressa nas primeiras sincronias das manifestações literárias da poesia escrita em língua portuguesa. / Universidade Federal da Bahia. Instituto de Letras. Salvador-Ba, 2009

The aim of this thesis is the identification and evaluation lexical expression of causal and consequential relations in the present-day French. After a theoretical introduction in which the most important terms necessary for this work will be presented, we will focus on the Czech conception of discourse relations. Based on this conception, a list of French lexical expressions of cause and consequence will be assembled. These expressions will be then searched in two corpora ? frWac and Frantext. In the practical part of this piece of writing, the given lexical expressions will be analysed with respect to their structure, position and use in a sentence and in the text. The frequency of their occurrence in both corpora will be also taken in account.

La lexicographie contemporaine met à disposition des ressources offrant de multiples possibilités d’exploitation automatique. Ainsi, le Réseau Lexical du Français, en cours d’élaboration, est un graphe monolingue, constitué de sommets, entre lesquels sont encodées des relations syntaxico-sémantiques. Cette thèse s’intéresse à son exploration par raisonnement analogique. Elle débute par une revue de la formalisation et de l’informatisation de l’analogie pour l’étude du lexique, qui définit les principes de l’exploration  : les sommets sont des objets disposant d’Attributs, les arcs représentent des Relations. Une réflexion est menée sur la nature de ces éléments et les rapports qu’ils entretiennent, réalisée en tenant compte leur évolution dans le temps et de la topologie du graphe. Deux séries d’expériences viennent ensuite. La première montre que la formalisation de la ressource permet de détecter des analogies conformes à l’intuition, que différents types d’exploration sont possibles et que l’approche permet de vérifier la cohérence du réseau et de faire émerger des règles lexicales. La seconde série porte sur la notion de configurations de dérivations lexicales. Elle montre que le regroupement de sous-graphes analogues fait émerger des connexions récurrentes. L’état d’avancement du réseau ne permet pas d’obtenir des règles et des modèles aboutis, mais les résultats sont encourageants. L’analogie est alors considérée comme un guide pour s’assurer de la qualité de la représentation du lexique proposée et acquérir des connaissances sur son organisation. Elle permet d’identifier des phénomènes linguistiques et d’instrumenter l’activité lexicologique. / Contemporary lexicography provide ressources offering many opportunities for natural language processing tasks. Thus, the French Lexical Network, presently under development, is a graph of lexical units connected by a rich set of lexical relations. This PhD thesis lays the groundwork for an exploration of this ressource by analogical reasonning. It begins with a selective overview of formalisation and computerisation for study of lexicon, wich defines the principle of exploration : the nodes are similar to objects, which have some attributes and edges represent relations. A reflection is conducted on the nature of this constituents and the relations between them. It takes into account the time axis and the topology of the network. Then two sets of exploratory experiments are conducted. The first one shows that the resource formalisation makes it possible to detect automatically analogies consistent with intuition, that several kind of analogical explorations are possible and that the approach allows to check the consistency of the resource and to bring out lexical rules. The second one is focused around the concept of lexical derivation configurations. It shows how grouping of analogous subgraphs reveals recurrent connections. The progress status of the resource doesn't enable us to obtain successfully completed rules and models, but results are nontheless encouraging. Analogy can already be considered as a guide to ensure the quality of lexical resources. It also allows for the acquisition of knowledge about its organisation. Such knowledge can be used to identify linguitic phenomena and to design instruments to support lexicographic activity.