Dans cette thèse, nous abordons du point de vue linguistique la question de l'évaluation des bases lexicales extraites par analyse distributionnelle automatique (ADA). Les méthodes d'évaluation de ces ressources qui sont actuellement mises en œuvre (comparaison à des lexiques de référence, évaluation par la tâche, test du TOEFL...) relèvent en effet d'une approche quantitative des données qui ne laisse que peu de place à l'interprétation des rapprochements générés. De ce fait, les conditions qui font que certains couples de mots sont extraits alors que d'autres ne le sont pas restent mal connues. Notre travail vise une meilleure compréhension des fonctionnements en corpus qui régissent les rapprochements distributionnels. Pour cela, nous avons dans un premier temps adopté une approche quantitative qui a consisté à comparer plusieurs ressources distributionnelles calculées sur des corpus différents à des lexiques de références (le Dictionnaire électronique des synonymes du CRISCO et le réseau lexical JeuxDeMots). Cette étape nous a permis, premièrement, d'avoir une estimation globale du contenu de nos ressources, et, deuxièmement, de sélectionner des échantillons de couples de mots à étudier d'un point de vue qualitatif. Cette deuxième étape constitue le cœur de la thèse. Nous avons choisi de nous focaliser sur les relations lexico-sémantiques que sont la synonymie, l'antonymie, l'hyperonymie et la méronymie, que nous abordons en mettant en place quatre protocoles différents. En nous appuyant sur les relations contenues dans les lexiques de référence, nous avons comparé les propriétés distributionnelles des couples de synonymes/antonymes/hyperonymes/méronymes qui ont été extraits par l'ADA avec celles des couples qui ne l'ont pas été. Nous mettons ainsi au jour plusieurs phénomènes qui favorisent ou bloquent la substituabilité des couples de mots (donc leur extraction par l'ADA). Ces phénomènes sont considérés au regard de paramètres comme la nature du corpus qui a permis de générer les bases distributionnelles étudiées (corpus encyclopédique, journalistique ou littéraire) ou les limites des lexiques de référence. Ainsi, en même temps qu'il questionne les méthodes d'évaluation des bases distributionnelles actuellement employées, ce travail de thèse illustre l'intérêt qu'il y a à considérer ces ressources comme des objets d'études linguistiques à part entière. Les bases distributionnelles sont en effet le résultat d'une mise en œuvre à grande échelle du principe de substituabilité, ce qui en fait un matériau de choix pour la description des relations lexico-sémantiques. / In this thesis, we address the question of the evaluation of distributional thesauri from a linguistic point of view. The most current ways to evaluate distributional methods rely on the comparison with gold standards like WordNet or semantic tasks like the TOEFL test. However, these evaluation methods are quantitative and thus restrict the possibility of performing a linguistic analysis of the distributional neighbours. Our work aims at a better understanding of the distributional behaviors of words in texts through the study of distributional thesauri. First, we take a quantitative approach based on a comparison of several distributional thesauri with gold standards (the DES - a dictionary of synonyms - and JeuxDeMots - a crowdsourced lexical network). This step allowed us to have an overview of the nature of the semantic relations extracted in our distributional thesauri. In a second step, we relied on this comparison to select samples of distributional neighbours for a qualitative study. We focused on "classical" semantic relations, e.g. synonymy, antonymy, hypernymy and meronymy. We considered several protocols to compare the properties of the couples of distributional neighbours which were found in the gold standards and the others. Thus, taking into account parameters like the nature of the corpora from which were generated our distributional thesauri, we explain why some synonyms, hypernyms, etc. can be substituted in texts while others cannot. The purpose of this work is twofold. First, it questions the traditional evaluation methods, then it shows how distributional thesauri can be used for the study of semantic relations.

Several theorists (amongst whom Halliday & Matthiessen, 2004:175) have stated that the verb (or the verbal group) is the core element in clauses and is largely responsible for the grammatical structure of language. In spite of this pivotal role that verbs fulfil in clauses, lexical specificity is rarely investigated in this class. Instead researchers prefer to investigate lexical specificity in the noun class. It is against this background that the main purpose of this study is to investigate specificity in the lexical verbs of first language (L1) English users and Black South African English (BSAE) users. In order to achieve this aim the secondary aims of the study are: to develop a framework for the analysis of lexical specificity in the lexical verb word class (ii) to compare specificity with regard to lexical verbs in L1 English and BSAE by using corpora, which allows one to firstly test the analysis framework on corpus data and secondly to determine whether or not lexical verbs in BSAE are less specific than lexical verbs in L1 English (iii) to establish the reasons for the differences in lexical specificity with regard to lexical verbs between L1 English and BSAE English. In order to achieve these aims, two corpora were used: the Louvain Corpus of Native English Essays (also known as LOCNESS) that consists of texts written by a subpopulation of L1 English users and the Tswana Learner English Corpus (also known as the TLE) that consists of texts written by a subpopulation of BSAE users. The results obtained in this study were interpreted from both a systemic functional perspective and a cognitive perspective. The initial quantitative results indicated that even though the BSAE users use lexical verbs (tokens) more frequently than L1 English users do, the BSAE users have fewer lexical verb lemmas (types) at their disposal than the L1 English users. Statistical tests determined that the mean type/token ratio between the two independent corpora is not only significant, but that the degree to which there is more specificity in the lexical verbs of the L1 English users than in the lexical verbs of the BSAE users is large. Due to space constraints, the qualitative part of the study focused only on the communication verbs in LOCNESS and the TLE. In order to provide an in-depth overview of the communication verbs, the communication verbs were divided into five semantic subcategories (using frequency counts and semantic considerations). They are: (i) to say something in a particular manner] (ii) to say something in order to express one’s feelings] (iii) to say something in order to convey information] (iv) to say something to someone in order to elicit a certain response] (v) to say something in response to something already said] Each of these semantic categories in turn was divided into semantic subcategories to present a detailed insight into the communication verbs employed in both LOCNESS and the TLE. In the study it was determined that there is overall more specificity in the lexical verbs of the L1 English users than in the lexical verbs of the L2 English users. It was also determined that there is overall more lexical diversity within the lexical verbs of the L1 English users. The following factors influenced the lexical specificity and lexical diversity in the corpora: (i) In cases where more general communication verbs are elaborated by verbs containing manner elaborations, the BSAE users tend not to use the more specific verbs. (ii) Communication verbs that usually play an important role in academic literacy (such as summarise and argue) are used to a lesser extent by the BSAE users than the L1 English users. (iii) In cases where a communication verb could possibly belong to another semantic category (as is the case with the verbs stress and maintain) the BSAE users tend to avoid using the verbs as communication verbs. (iv) Some communication verbs (such as demand and beg) acquired additional meanings in BSAE. (v) Some essay topics in both LOCNESS and the TLE influenced the frequencies of some communication verbs in the respective corpora. Considering the factors above, it was found that being a L2 English user means that English will not always function in the same contexts for the BSAE users as it would for the L1 English users. Therefore the vocabulary of the BSAE users will only be specific and diversified in those semantic categories needed to function in certain contexts. Consequently, the findings of this study can be used to contribute to the development of pedagogical material in academic literacy courses being presented to BSAE users to create an awareness of the variation in English and all the contexts in which it can function. The findings can also be of value to researchers in the fields of lexicography and computational linguistics. / Thesis (M.A. (English))--North-West University, Vaal Triangle Campus, 2011

Several theorists (amongst whom Halliday & Matthiessen, 2004:175) have stated that the verb (or the verbal group) is the core element in clauses and is largely responsible for the grammatical structure of language. In spite of this pivotal role that verbs fulfil in clauses, lexical specificity is rarely investigated in this class. Instead researchers prefer to investigate lexical specificity in the noun class. It is against this background that the main purpose of this study is to investigate specificity in the lexical verbs of first language (L1) English users and Black South African English (BSAE) users. In order to achieve this aim the secondary aims of the study are: to develop a framework for the analysis of lexical specificity in the lexical verb word class (ii) to compare specificity with regard to lexical verbs in L1 English and BSAE by using corpora, which allows one to firstly test the analysis framework on corpus data and secondly to determine whether or not lexical verbs in BSAE are less specific than lexical verbs in L1 English (iii) to establish the reasons for the differences in lexical specificity with regard to lexical verbs between L1 English and BSAE English. In order to achieve these aims, two corpora were used: the Louvain Corpus of Native English Essays (also known as LOCNESS) that consists of texts written by a subpopulation of L1 English users and the Tswana Learner English Corpus (also known as the TLE) that consists of texts written by a subpopulation of BSAE users. The results obtained in this study were interpreted from both a systemic functional perspective and a cognitive perspective. The initial quantitative results indicated that even though the BSAE users use lexical verbs (tokens) more frequently than L1 English users do, the BSAE users have fewer lexical verb lemmas (types) at their disposal than the L1 English users. Statistical tests determined that the mean type/token ratio between the two independent corpora is not only significant, but that the degree to which there is more specificity in the lexical verbs of the L1 English users than in the lexical verbs of the BSAE users is large. Due to space constraints, the qualitative part of the study focused only on the communication verbs in LOCNESS and the TLE. In order to provide an in-depth overview of the communication verbs, the communication verbs were divided into five semantic subcategories (using frequency counts and semantic considerations). They are: (i) to say something in a particular manner] (ii) to say something in order to express one’s feelings] (iii) to say something in order to convey information] (iv) to say something to someone in order to elicit a certain response] (v) to say something in response to something already said] Each of these semantic categories in turn was divided into semantic subcategories to present a detailed insight into the communication verbs employed in both LOCNESS and the TLE. In the study it was determined that there is overall more specificity in the lexical verbs of the L1 English users than in the lexical verbs of the L2 English users. It was also determined that there is overall more lexical diversity within the lexical verbs of the L1 English users. The following factors influenced the lexical specificity and lexical diversity in the corpora: (i) In cases where more general communication verbs are elaborated by verbs containing manner elaborations, the BSAE users tend not to use the more specific verbs. (ii) Communication verbs that usually play an important role in academic literacy (such as summarise and argue) are used to a lesser extent by the BSAE users than the L1 English users. (iii) In cases where a communication verb could possibly belong to another semantic category (as is the case with the verbs stress and maintain) the BSAE users tend to avoid using the verbs as communication verbs. (iv) Some communication verbs (such as demand and beg) acquired additional meanings in BSAE. (v) Some essay topics in both LOCNESS and the TLE influenced the frequencies of some communication verbs in the respective corpora. Considering the factors above, it was found that being a L2 English user means that English will not always function in the same contexts for the BSAE users as it would for the L1 English users. Therefore the vocabulary of the BSAE users will only be specific and diversified in those semantic categories needed to function in certain contexts. Consequently, the findings of this study can be used to contribute to the development of pedagogical material in academic literacy courses being presented to BSAE users to create an awareness of the variation in English and all the contexts in which it can function. The findings can also be of value to researchers in the fields of lexicography and computational linguistics. / Thesis (M.A. (English))--North-West University, Vaal Triangle Campus, 2011

L’hypothèse sur laquelle repose notre travail est que la comparaison de la lexicalisation des noms d’éléments du corps (dorénavant, NEC) et de la phraséologie des NEC entre deux langues va permettre de mettre en évidence des différences de conceptualisation et de culture entre deux sociétés. En fonction de cette hypothèse, notre thèse aborde deux thèmes principaux. Premièrement, nous étudions les NEC coréens (dorénavant, NECC) en nous focalisant sur les noms neutres d’éléments externes du corps humain. Les NECC ont des caractéristiques universelles : richesse lexicale, éléments du vocabulaire basique, source de l’« embodiment », universel physio-conceptuel et nature de quasi-prédicats sémantiques. En même temps, les NECC montrent des particularités sémantiques, syntaxiques et morphologiques liées aux spécificités de la langue coréenne. La comparaison de la lexicalisation des NECC et des NEC français montre que même si les éléments du corps sont des universels physio-conceptuels, il n’y a pas de correspondance lexicale univoque entre les deux langues. Deuxièmement, nous focalisons notre attention sur la phraséologie des NECC et sa modélisation dans le Réseau Lexical du Coréen, une modélisation lexicographique formelle fondée sur une conception relationnelle du lexique. Nous bornons la phraséologie des NECC aux collocations contrôlées par les NECC (par ex. koga oddukhada, litt. nez+SUB être.haut ‘avoir un nez haut et beau’). Dans la phraséologie des NECC, nous prenons aussi en compte la phraséologisation dans un mot-forme (par ex. napjakko, ‘nez aplati’). Nous appelons collocation morphologisée ce type de phrasème morphologique par opposition à la collocation lexicale. À partir de l’examen des collocations non seulement lexicales mais aussi morphologisées contrôlées par NEC, nous pouvons obtenir les composantes sémantiques de la définition de la base, le NEC. Après cela, nous proposons un patron universel de définition des NEC, qui est le fondement du modèle explicatif de la phraséologie des NEC. Ce modèle s’appuie sur l’hypothèse selon laquelle on peut trouver dans les définitions des NEC des composantes récurrentes. Différentes collocations (du type Magn, Ver, Bon, Real1, en termes de fonctions lexicales de la Théorie Sens-Texte) sont alors générées relativement au sémantisme de ces composantes. Finalement, nous comparons la description de la phraséologie des NECC à celle des NEC français, afin d’observer les diverses non-correspondances entre les phrasèmes des deux langues. Ce travail approfondit notre compréhension de la phraséologie aussi bien en général, qu’en tant qu’elle est appliquée au coréen et au français, et met en relief des différences culturelles encodées dans les deux langues. Il peut également trouver des applications en didactique et en traductologie. / The hypothesis on which our work is based is that the comparison of the lexicalization of body element nouns (henceforth, NEC, Fr. nom d’élément du corps) and the phraseology of the NEC between two languages will make it possible to highlight the differences of conceptualization and culture between two societies.According to this hypothesis, our thesis deals with two main themes. Firstly, we study the Korean NEC (henceforth, NECC, Fr. nom d’élément du corps coréen) focusing on the neutral nouns of human external body elements. The NECC have universal characteristics: lexical richness, elements of the basic vocabulary, sources of the embodiment, physio-conceptual universals and their nature of semantic quasi-predicates. At the same time, the NECC show language-specific semantic, syntactic and morphological characteristics. The comparison of the lexicalization of the NECC and the French NEC shows that even if the elements of the body are physio-conceptual universals, there is no univocal lexical correspondence between the two languages.Secondly, we focus our attention on the NECC’s phraseology and its modeling in the Korean Lexical Network, a formal lexicographic model based on a relational conceptualization of the lexicon. We limit the NECC’s phraseology to collocations the NECC control (ex. koga oddukhada, ‘have a high and pretty nose’). Within the NECC’s phraseology, we also take into account the phraseologisation in a word-form (ex. napjakko, ‘flat nose’). We denote this morphological phraseme by the term morphologised collocation, as opposed to the lexical collocation. From the examination of lexical and morphologised collocations which NECC control, we can identify the semantic components of the definition of the NECC. After that, we propose a universal definition pattern of the NEC, which is the foundation of the explanatory model of the NEC’s phraseology. This model is based on the assumption that recurrent components can be found in the definitions of NEC. Different collocations (of the type Magn, Ver, Bon, Real1, in terms of Lexical Functions of the Meaning-Text Theory) are then generated from the semantism of these components. Finally, we compare the description of the phraseology of the NECC with that of the French NEC, in order to observe the various non-correspondences between the phrasemes of the two languages.This work deepens our understanding of phraseology in general and in specific languages (Korean and French), and highlights cultural differences encoded in both languages. It can also find applications in didactics and translation.

Submitted by Renata Lopes (renatasil82@gmail.com) on 2016-09-09T12:38:29Z No. of bitstreams: 1 danielpereiraalves.pdf: 2032305 bytes, checksum: 7b4aa5be0177e28263a5fdb95d14c75e (MD5) / Approved for entry into archive by Diamantino Mayra (mayra.diamantino@ufjf.edu.br) on 2016-09-13T15:13:04Z (GMT) No. of bitstreams: 1 danielpereiraalves.pdf: 2032305 bytes, checksum: 7b4aa5be0177e28263a5fdb95d14c75e (MD5) / Made available in DSpace on 2016-09-13T15:13:04Z (GMT). No. of bitstreams: 1 danielpereiraalves.pdf: 2032305 bytes, checksum: 7b4aa5be0177e28263a5fdb95d14c75e (MD5) Previous issue date: 2010-08-20 / CAPES - Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / Este trabalho investiga o papel de fronteiras de constituintes prosódicos no processamento lexical de falantes adultos do Português Brasileiro (PB). Assume-se que a fala é organizada em uma hierarquia de constituintes prosódicos que fazem interface com outros constituintes gramaticais, ainda que essa relação não seja necessariamente biunívoca (NESPOR & VOGEL, 1986). A partir de evidências experimentais sobre o papel da informação prosódica no processamento sintático e lexical (Christophe et. al., 2004; Millote et al., 2008; Silva & Name, 2009), propõem-se dois experimentos com o intuito de se investigar a ação de fronteiras de constituintes prosódicos no acesso lexical on-line, por meio da técnica de detecção de palavras. O primeiro experimento analisou o efeito da ambigüidade local (gol/ golfe) presente em fronteiras de palavra prosódica (FPP): [gol ω final]. O segundo experimento examinou o mesmo efeito em fronteiras de sintagma fonológico (FSF): [...gol] ϕ [ficou...]. Os resultados indicam um efeito significativo da ambiguidade em FPP, e não em FSF. Quer isso dizer que as informações acústicas em FSF atuaram na inibição de itens lexicais potencialmente competidores (golfe) no PB. Discutem-se as implicações de tal resultado para o processamento da linguagem à luz de modelos de modelos de acesso lexical (LEVELT, 1989; MCCLELLAND & ELMAN, 1986), do modelo de computação linguística do Programa Minimalista (HAUSER, CHOMSKY e FITCH, 2002; CHOMSKY, 1999) e da possibilidade de integração entre teoria linguística e modelos de processamento de linguagem (CÔRREA & AUGUSTO, 2006). / This study investigates the role of prosodic constituents’ boundaries in lexical processing by Brazilian Portuguese (BP) adult speakers. It is assumed that speech is organized in a hierarchy of prosodic constituents that interface with other grammatical constituents, although this relation is not necessarily mapped in a one-to-one fashion (NESPOR & VOGEL, 1986). Motivated by experimental evidence on the role of prosodic information in syntactic and lexical processing (CHRISTOPHE et al., 2004, MILLOTE et al., 2008; SILVA & NAME, 2009), two experiments are proposed in order to investigate the role of prosodic constituents’ boundaries in on-line lexical access by means of word monitoring tasks. The first experiment analyzed the local ambiguity (gol/golfe) effect on prosodic word boundaries (FPP): [gol ω final]. The second experiment examined the same effect on phonological phrase boundaries (FSF): [...gol] ϕ [ficou...]. The results indicate a significant ambiguity effect in the FPP condition and not in the FSF one. This means that the acoustic information in FSF played a significant role in the inhibition of competitor lexical items (golfe) in BP. The implications of this result for language processing are discussed in light of lexical access models (LEVELT, 1989; MCCLELLAND & ELMAN, 1986), the language computation model of the Minimalist Program (HAUSER, CHOMSKY e FITCH, 2002; CHOMSKY, 1999) and the possibility of integration between linguistic theory and language processing models (CÔRREA & AUGUSTO, 2006).

published_or_final_version / Linguistics / Doctoral / Doctor of Philosophy

Computer Sciences / In order to respond to increasing demand for natural language interfaces---and provide meaningful insight into user query intent---fast, scalable lexical semantic models with flexible representations are needed. Human concept organization is a rich phenomenon that has yet to be accounted for by a single coherent psychological framework: Concept generalization is captured by a mixture of prototype and exemplar models, and local taxonomic information is available through multiple overlapping organizational systems. Previous work in computational linguistics on extracting lexical semantic information from unannotated corpora does not provide adequate representational flexibility and hence fails to capture the full extent of human conceptual knowledge. In this thesis I outline a family of probabilistic models capable of capturing important aspects of the rich organizational structure found in human language that can predict contextual variation, selectional preference and feature-saliency norms to a much higher degree of accuracy than previous approaches. These models account for cross-cutting structure of concept organization---i.e. selective attention, or the notion that humans make use of different categorization systems for different kinds of generalization tasks---and can be applied to Web-scale corpora. Using these models, natural language systems will be able to infer a more comprehensive semantic relations, which in turn may yield improved systems for question answering, text classification, machine translation, and information retrieval. / text

Notre motivation est de combler le fossé terminologique qui grandit avec la production massive de nouveaux concepts (50 quotidiens) dans divers domaines, pour lesquels les termes sont souvent inventés d'abord dans une certaine langue bien dotée, telle que l'anglais ou le français. Trouver des termes équivalents dans différentes langues est nécessaire pour de nombreuses applications, telles que la RI translingue et la TA. Cette tâche est très difficile, particulièrement pour certaines langues très utilisées telles que l'arabe, parce que (1) seule une petite proportion de nouveaux termes est correctement enregistrée par des terminologues, et pour peu de langues ; (2) des communautés spécifiques créent continuellement des termes équivalents sans les normaliser ni même les enregistrer (terminologie latente) ; (3) dans de nombreux cas, aucuns termes équivalents ne sont créés, formellement ou informellement (absence de terminologie). Cette thèse propose de remplacer le but impossible de construire d'une manière continue une terminologie à jour, complète et de haute qualité pour un grand nombre de langues par celui de construire une preterminologie, en utilisant des méthodes non conventionnelles et des contributions passives ou actives par des communautés d'internautes : extraction de termes parallèles potentiels non seulement à partir de textes parallèles ou comparables, mais également à partir des logs (traces) des visites à des sites Web tels que DSR (Route de la Soie Digitale), et à partir de données produites par des jeux sérieux. Une préterminologie est un nouveau genre de ressource lexicale qui peut être facilement construit et a une bonne couverture. Suivant en ceci une tendance croissante en lexicographie computationnelle et en TALN en général, nous représentons une préterminologie multilingue par une structure de graphe (Preterminological Multilingual Graph, MPG), où les nœuds portent des prétermes et les arcs des relations préterminologiques simples (synonymie monolingue, traduction, généralisation, spécialisation, etc.) qui sont des approximations des relations (terminologiques ou ontologiques) usuelles. Un Système complet pour Éliciter une Préterminologie (SEPT) a été développé pour construire et maintenir des MPG. Des approches passives ont été expérimentées en développant un MPG pour le site Web culturel de DSR, et un autre pour le domaine de l'onirologie arabe : les ressources produites ont atteint une bonne couverture informationnelle et linguistique. L'approche indirecte par contribution active est testée depuis 8-9 mois sur l'instance arabe du jeu sérieux JeuxDeMots.

No description available.

No description available.