Academic literature on the topic 'Syntactic constituents'

Made available in DSpace on 2017-07-10T18:56:11Z (GMT). No. of bitstreams: 1 OLIZAROSKI_ Holland.pdf: 2752438 bytes, checksum: 8d1894b2013ac1f1db78f27e44e68a2c (MD5) Previous issue date: 2017-02-03<br>Between many inquiries that have been challenging the scientific community from a grammatical point of view, one, to which there is no consent yet, is related to the order of the constituents of the sentences produced in the Brazilian Sign Language (Libras), because, although it is the language of deaf people in Brazil since 2002, official by the means of the Law nº 10.436 and regulated through the decree nº 5.626 in 2005, one of the big problems for its use and linguistic description refers to the little knowledge that oneself has regarding the organization and structuring of its sentences. Thus, discussing the Libras in a linguistic-theoretical perspective, some syntactic problems emerge, which bring many inquiries, being three of them the ones that guide this research, namely: (i). Which are the syntactic pattern accepted in Libras? (ii) What motivates and/or licenses these patterns? And; (iii) The syntactic orders shown in Libras sentences would be, exclusively, due to its visuospatial modality, with the kind of verb that can be shown? Thus, with the purpose of finding answers to this inquiry, we traced, as the main objective, the reflection over the syntagm organization in sentences produced in Libras . On the perspective of reaching this objective, we based the research on the theoretical assumptions of the Functional Linguistic, which tries to explain the sentence structure according its linguistic function. Being this way, we start from the Greenberg&#8223;s (1963) postulates, to whom the great majority of languages have diverse variable orders, but only one as dominant, and they can be distributed in SVO, SOV, VSO, VOS, OSV or OVS: we passed, among others, by Hopper and Thompson (1980), with the intention of verifying if the conceived transitivity by the means of a continuum scale of ten parameters influence in the syntactic constituent organization of sentences in Libras; we sought, as well, in Chafe (1979), Borba (2002) and Ferreira Brito (2010) punctual inquiries referring to verb with the syntactic-semantic value and the visuospatial modality classification. In this perspective, as methodology we adopted the basic research, the bibliographical revision type and the qualitative nature. As technic and procedure of collecting data the selection of sentences in Portuguese Language in the Portuguese Corpus/2006, submitting them, after their preparation, to a deaf informer interpretation, which has remitted us to a field research. The transition of these sentences to gloss-Libras resulted in a Parallel Corpus constituted of 114 sentences, by means of which we could perform syntactic analyses observing the syntagm organization of the sentences produced in Libras our main objective. As result of this research, we have verified that, in Libras, the SVO, SOV and OSV patterns tend to manifest themselves, being the transitivity a strong motivation and/or licensing motivation for these patterns, as well as the visuospatial modality with the syntactic-semantic value of the verb, because we realized, during the reflections, that the same kind of verb in classification in Libras can present distinct orders in the same transitivity pole, which is, low transitivity sentences, containing process verbs or state verbs and, in Libras, non-directional based to the body, tend, more commonly, to present the SVO pattern. On the other hand, the high transitivity sentences containing action-process and action verbs and, in Libras, irreversible directional, reversible directional, classifiers, non-directional based on the body, that incorporate the object and instrumental tend to present the SVO, SOV and OSV patterns. Thus, despite the disparity in classification in Libras these verbs coincide in syntactic-semantic value as well as in the sentential transitivity pole. This led us to deduce that only the kind of verb in its visuospatial modality cannot be preponderant motivator of the syntactic order constituents of sentences produced in Libras<br>Das muitas questões que vem desafiando a comunidade científica de um ponto de vista gramatical, uma, para a qual ainda não há consenso, diz respeito à ordem dos constituintes das sentenças produzidas na Língua Brasileira de Sinais (Libras), pois, embora seja ela língua das pessoas surdas do Brasil desde 2002, oficializada por meio da Lei nº 10.436 e regulamentada pelo Decreto nº 5.626 em 2005, um dos grandes problemas para sua utilização e descrição linguística refere-se ao pouco conhecimento que se tem quanto à organização e estruturação de suas sentenças. Assim, ao se discutir a Libras numa perspectiva linguístico-teórica, surgem problemas de cunho sintático, os quais acenam para várias indagações, sendo três delas as que nortearam essa pesquisa, a saber: (i). Quais são os padrões sintáticos admitidos pela Libras? (ii) O que motiva e/ou licencia esses padrões? e; (iii) As ordens sintáticas manifestadas nas sentenças produzidas em Libras seria, exclusivamente, em decorrência de sua modalidade visuoespacial, diante do tipo de verbo que pode apresentar? Assim, no propósito de encontrar resposta(s) a essa problematização, traçamos, como objetivo geral, a reflexão sobre a organização dos sintagmas das sentenças produzidas em Libras . Na perspectiva de alcançar esse objetivo, sustentamos a pesquisa nos pressupostos teóricos da Linguística Funcional, a qual tenta explicar a estrutura da sentença em termos de função linguística. Assim sendo, partimos de postulados de Greenberg (1963), para o qual a grande maioria das línguas tem diversas ordens variantes, mas apenas uma dominante, podendo ser elas distribuídas em SVO, SOV, VSO, VOS, OSV ou OVS; perpassamos, dentre outros, por Hopper e Thompson (1980), no intuito de verificar se a transitividade concebida por meio de um continuum escalar de dez parâmetros influencia na organização dos constituintes sintáticos das sentenças em Libras; buscamos, ainda, em Chafe (1979), Borba (2002) e Ferreira Brito (2010) questões pontuais referente ao verbo como o valor sintático-semântico e a classificação em modalidade visuoespacial. Nessa perspectiva, adotamos como metodologia a pesquisa de natureza básica, do tipo revisão bibliográfica e de cunho qualiquantitativo. Assumimos como técnica e procedimento de coleta de dados a seleção de sentenças em Língua Portuguesa no Corpus do Português/2006, submetendo-as, após sua preparação, à interpretação por um informante surdo, o que nos remeteu também à pesquisa de campo. A transcrição dessas sentenças para a glosa-Libras resultou em um Corpus Paralelo constituído de 114 sentenças, por meio do qual pudemos realizar análises sintáticas com vistas à reflexão sobre a organização dos sintagmas das sentenças produzidas em Libras nosso objetivo central. Como resultado desse processo de investigação, constatamos que, em Libras, tendem a manifestar-se os padrões SVO, SOV e OSV, sendo a transitividade forte indício de motivação e/ou licenciamento desses padrões, bem como a modalidade visuoespacial associada ao valor sintático-semântico do verbo, isso porque atinamos, no decorrer das reflexões, que o mesmo tipo de verbo em classificação na Libras pode apresentar ordens distintas num mesmo pólo de transitividade, ou seja, sentenças de baixa transitividade, contendo verbos de processo ou verbos de estado e, na Libras, não-direcionais ancorados ao corpo, tendem, mais comumente, a apresentar o padrão SVO. Já as sentenças de alta transitividade contendo verbos de ação-processo e ação e, na Libras, direcionais irreversíveis, direcionais reversíveis, classificadores, não-direcionais ancorados ao corpo, que incorporam o objeto e instrumentais tendem a apresentar os padrões SVO, SOV e OSV. Assim, apesar da disparidade em classificação na Libras esses verbos coincidem em valor sintático-semântico bem como no pólo de transitividade sentencial. Isso nos levou a deduzir que apenas o tipo de verbo em sua modalidade visuoespacial não pode ser preponderante motivador da ordem dos constituintes sintáticos das sentenças produzidas em Libras

L’analyse syntaxique consiste à prédire la représentation syntaxique de phrases en langue naturelle sous la forme d’arbres syntaxiques. Cette tâche pose des problèmes particuliers pour les langues non-configurationnelles ou qui ont une morphologie flexionnelle plus riche que celle de l’anglais. En particulier, ces langues manifestent une dispersion lexicale problématique, des variations d’ordre des mots plus fréquentes et nécessitent de prendre en compte la structure interne des mots-formes pour permettre une analyse syntaxique de qualité satisfaisante. Dans cette thèse, nous nous plaçons dans le cadre de l’analyse syntaxique robuste en constituants par transitions. Dans un premier temps, nous étudions comment intégrer l’analyse morphologique à l’analyse syntaxique, à l’aide d’une architecture de réseaux de neurones basée sur l’apprentissage multitâches. Dans un second temps, nous proposons un système de transitions qui permet de prédire des structures générées par des grammaires légèrement sensibles au contexte telles que les LCFRS. Enfin, nous étudions la question de la lexicalisation de l’analyse syntaxique. Les analyseurs syntaxiques en constituants lexicalisés font l’hypothèse que les constituants s’organisent autour d’une tête lexicale et que la modélisation des relations bilexicales est cruciale pour désambiguïser. Nous proposons un système de transition non lexicalisé pour l’analyse en constituants discontinus et un modèle de scorage basé sur les frontières de constituants et montrons que ce système, plus simple que des systèmes lexicalisés, obtient de meilleurs résultats que ces derniers<br>Syntactic parsing consists in assigning syntactic trees to sentences in natural language. Syntactic parsing of non-configurational languages, or languages with a rich inflectional morphology, raises specific problems. These languages suffer more from lexical data sparsity and exhibit word order variation phenomena more frequently. For these languages, exploiting information about the internal structure of word forms is crucial for accurate parsing. This dissertation investigates transition-based methods for robust discontinuous constituency parsing. First of all, we propose a multitask learning neural architecture that performs joint parsing and morphological analysis. Then, we introduce a new transition system that is able to predict discontinuous constituency trees, i.e.\ syntactic structures that can be seen as derivations of mildly context-sensitive grammars, such as LCFRS. Finally, we investigate the question of lexicalization in syntactic parsing. Some syntactic parsers are based on the hypothesis that constituent are organized around a lexical head and that modelling bilexical dependencies is essential to solve ambiguities. We introduce an unlexicalized transition system for discontinuous constituency parsing and a scoring model based on constituent boundaries. The resulting parser is simpler than lexicalized parser and achieves better results in both discontinuous and projective constituency parsing

Cette recherche a deux objectifs, la systématisation de la prosodie fondamentale du japonais langue commune et l’application de ce système prosodique à une méthode didactique. Pour atteindre ces objectifs, nous cernons d'abord un cadre théorique adaptable au japonais en nous fondant sur la théorie de la "Grammaire de l’intonation" de Morel et Danon-Boileau (1998). Ensuite, nous analysons deux types de corpus, les premières phrases d'informations télévisées, dont le style est proche de celui de l’écrit, et des extraits de dialogue spontané. Les résultats de l’analyse du corpus d'écrit oralisé corroborent une hypothèse selon laquelle l’unité discursive, qui est dans l’ordre fondamental des constituants tel que défini par la fonction de détermination correspondant à un marqueur syntaxique/discursif spécifique, doit être réalisée par la prosodie fondamentale conformément à la fonction de détermination. Les indices suprasegmentaux de cette prosodie sont la position et la durée de la pause et la mélodie des séquences du modifiant et du modifié. En application de ce système prosodique, nous proposons une méthode d’enseignement d’accès facile pour l’enseignant et l’apprenant sur les plans théorique et didactique, grâce au critère binaire de la fonction de détermination d'une part, et d’un indice "écrit", le marqueur syntaxique/discursif, d'autre part. L’analyse du corpus d’oral montre que dans l’oral la fonction de détermination n’est pas représentée par le marqueur spécifique ni par la pause, qui sont remplacés par un indice intonatif à la fin du constituant, tandis que la fonction énonciative est réalisée par l’intonation des particules finales<br>This research has two objectives: a systematization of the fundamental prosody of Japanese as a common language, and an application of this prosodic system to teaching method. To attain these objectives, we first fix a theoretical framework adaptable to Japanese based on the theory "Grammaire de l’intonation" (Intonation Grammar) of Morel and Danon-Boileau (1998). We then analyze two types of corpus, readings of lead sentences of TV news, whose style are close to the written style, and extracts from spontaneous dialogues. The results of analysis of the corpus in uttered-written style corroborate the hypothesis that the discursive unit, which is in the fundamental constituent order defined by the function of determination corresponding to a specific syntactic/discursive marker, must be realized by the fundamental prosody conforming to the function of determination. The supra-segmental indicia of this prosody are the position and the length of pauses and the pitch pattern of sequences of modifier and modified. With application of this prosodic system, we propose a teaching method easily accessible both to the teacher and the learner in theoretical and didactical aspects through the binary criterion of the function of determination and a "written" indicium, the specific syntactic/discursive marker. The analysis of the corpus in spoken style shows that in this style, the function of determination is represented neither by the marker nor by the pause, which are replaced by the intonated indicium at the end of constituent, whereas the enunciative function is realized by intonation of the final particles

Thesis (Ph. D.)--University of California, Los Angeles, 1988.<br>Typescript. Vita. eContent provider-neutral record in process. Description based on print version record. Includes bibliographical references (leaves 290-298).

Cette thèse par article comprend quatre articles qui contribuent au domaine de l'apprentissage profond, en particulier à l'accélération de l’apprentissage par le biais de réseaux à faible précision et à l'application de réseaux de neurones profonds au traitement du langage naturel. Dans le premier article, nous étudions un schéma d’entraînement de réseau de neurones qui élimine la plupart des multiplications en virgule flottante. Cette approche consiste à binariser ou à ternariser les poids dans la propagation en avant et à quantifier les états cachés dans la propagation arrière, ce qui convertit les multiplications en changements de signe et en décalages binaires. Les résultats expérimentaux sur des jeux de données de petite à moyenne taille montrent que cette approche produit des performances encore meilleures que l’approche standard de descente de gradient stochastique, ouvrant la voie à un entraînement des réseaux de neurones rapide et efficace au niveau du matériel. Dans le deuxième article, nous avons proposé un mécanisme structuré d’auto-attention d’enchâssement de phrases qui extrait des représentations interprétables de phrases sous forme matricielle. Nous démontrons des améliorations dans 3 tâches différentes: le profilage de l'auteur, la classification des sentiments et l'implication textuelle. Les résultats expérimentaux montrent que notre modèle génère un gain en performance significatif par rapport aux autres méthodes d’enchâssement de phrases dans les 3 tâches. Dans le troisième article, nous proposons un modèle hiérarchique avec graphe de calcul dynamique, pour les données séquentielles, qui apprend à construire un arbre lors de la lecture de la séquence. Le modèle apprend à créer des connexions de saut adaptatives, ce qui facilitent l'apprentissage des dépendances à long terme en construisant des cellules récurrentes de manière récursive. L’entraînement du réseau peut être fait soit par entraînement supervisée en donnant des structures d’arbres dorés, soit par apprentissage par renforcement. Nous proposons des expériences préliminaires dans 3 tâches différentes: une nouvelle tâche d'évaluation de l'expression mathématique (MEE), une tâche bien connue de la logique propositionnelle et des tâches de modélisation du langage. Les résultats expérimentaux montrent le potentiel de l'approche proposée. Dans le quatrième article, nous proposons une nouvelle méthode d’analyse par circonscription utilisant les réseaux de neurones. Le modèle prédit la structure de l'arbre d'analyse en prédisant un scalaire à valeur réelle, soit la distance syntaxique, pour chaque position de division dans la phrase d'entrée. L'ordre des valeurs relatives de ces distances syntaxiques détermine ensuite la structure de l'arbre d'analyse en spécifiant l'ordre dans lequel les points de division seront sélectionnés, en partitionnant l'entrée de manière récursive et descendante. L’approche proposée obtient une performance compétitive sur le jeu de données Penn Treebank et réalise l’état de l’art sur le jeu de données Chinese Treebank.<br>This thesis by article consists of four articles which contribute to the field of deep learning, specifically in the acceleration of training through low-precision networks, and the application of deep neural networks on natural language processing. In the first article, we investigate a neural network training scheme that eliminates most of the floating-point multiplications. This approach consists of binarizing or ternarizing the weights in the forward propagation and quantizing the hidden states in the backward propagation, which converts multiplications to sign changes and binary shifts. Experimental results on datasets from small to medium size show that this approach result in even better performance than standard stochastic gradient descent training, paving the way to fast, hardware-friendly training of neural networks. In the second article, we proposed a structured self-attentive sentence embedding that extracts interpretable sentence representations in matrix form. We demonstrate improvements on 3 different tasks: author profiling, sentiment classification and textual entailment. Experimental results show that our model yields a significant performance gain compared to other sentence embedding methods in all of the 3 tasks. In the third article, we propose a hierarchical model with dynamical computation graph for sequential data that learns to construct a tree while reading the sequence. The model learns to create adaptive skip-connections that ease the learning of long-term dependencies through constructing recurrent cells in a recursive manner. The training of the network can either be supervised training by giving golden tree structures, or through reinforcement learning. We provide preliminary experiments in 3 different tasks: a novel Math Expression Evaluation (MEE) task, a well-known propositional logic task, and language modelling tasks. Experimental results show the potential of the proposed approach. In the fourth article, we propose a novel constituency parsing method with neural networks. The model predicts the parse tree structure by predicting a real valued scalar, named syntactic distance, for each split position in the input sentence. The order of the relative values of these syntactic distances then determine the parse tree structure by specifying the order in which the split points will be selected, recursively partitioning the input, in a top-down fashion. Our proposed approach was demonstrated with competitive performance on Penn Treebank dataset, and the state-of-the-art performance on Chinese Treebank dataset.