Dissertations / Theses on the topic 'English adverb'

In the last few decades, there has been increased interest in the incorporation of data from bi- and multilingual individuals in linguistic theory: from second language acquisition and language attrition to heritage varieties and code-switching. This dissertation discusses a range of ways in which code-switching data can provide insight into the mechanisms that underlie linguistic structures. The data will be analysed within the framework of Minimalist Generative syntax and Distributed Morphology.The first part investigates grammatical gender assignment in code- switching between English, a language without grammatical gender, and two languages with grammatical gender: French and Belgian Dutch. These languages have comparable, but different gender systems. French has two genders: masculine and feminine, whereas Belgian Dutch adds a third: neuter. The study in this part of the dissertation compares gen- der assignment strategies in bilinguals with different profiles. In addition, the code-switching data provide evidence against the default status of neuter in Belgian Dutch.The second part focuses on word order and includes two studies: one on verb-second word order in Dutch-English code-switching and one on adverb placement in English-French and Dutch-English code- switching. The verb-second chapter identifies a lacuna in the traditional Generative analysis for verb second and uses the CS data to address this. The chapter on adverb position looks at placement of the adverb between the verb and its direct object, which is allowed in Dutch and French, but not in English. For all domains investigated, it is found that the finite verb predicts word order.Taken together, these studies demonstrate that bilingual data can shine a light on elements of the theory of grammar which remain in the shadows when only monolingual data is used.
Les dernières décennies ont vu croître l’intérêt pour l’intégration à la réflexion en linguistique théorique des données produites par des locuteurs/trices bilingues ou multilingues, que celles-ci concernent l’acquisition d’une langue seconde, l’attrition, les langues d’héritage ou l’alternance codique. Le présent travail développe plusieurs exemples où les données issues de l’alternance codique éclairent les mécanismes qui sous-tendent les structures linguistiques. Les données recueillies sont interprétées dans le cadre de la syntaxe générative minimaliste et de la morphologie distribuée (« distributed morphology »).Dans un premier temps, nous analysons l’attribution du genre grammatical dans l’alternance entre l’anglais, d’une part, et le français et le néerlandais de Belgique, de l’autre. Alors qu’il n’y a pas en anglais de genre grammatical, le français et le néerlandais de Belgique marquent ce genre, mais de façon différente :si le français distingue deux genres, masculin et féminin, le néerlandais de Belgique y adjoint un troisième, le neutre. Dans cette partie de la thèse, nous dressons le profil des stratégies d’attribution du genre auprès de deux types distincts de bilingues et nous établissons également que le neutre n’est pas le genre par défaut en néerlandais de Belgique.Dans un second temps, nous nous penchons sur l’ordre des constituants. Dans une première étude, nous examinons l’ordre des mots avec « verbe second » (V2) dans l’alternance anglais-néerlandais. Nous abordons ensuite le placement de l’adverbe dans l’alternance anglais- français et anglais-néerlandais. Le chapitre consacré à V2 identifie une lacune dans la littérature générative et tire profit des données de l’al- ternance pour y proposer une solution. Le chapitre consacré à l’adverbe s’intéresse au placement de celui-ci entre le verbe et son objet, position licite en français et néerlandais mais pas en anglais. Dans ces deux études, il apparaît que c’est la langue du verbe à la forme finie qui prédit l’ordre des constituants.L’ensemble des recherches ici réunies démontre que les données bilingues mettent en lumière des aspects de la théorie grammaticale qui restent dans l’ombre lorsque le chercheur se limite à des données monolingues.
Doctorat en Langues, lettres et traductologie
info:eu-repo/semantics/nonPublished

This thesis has attempted to determine if there are differences, concerning adverbial placement, between the sentences of left- and right-handed writers. To make this determination, I have statistically analyzed compositions of eight graduate students (four left-handed and four right-), and two left-handed published authors' (Lewis Carroll's and Mark Twains) private correspondence.

In this thesis I use the Principles and Parameters model of generative grammar to explain the surface distribution of adverbs in English. Using the current parameters and principals assumed for UG plus the rule of move $\alpha,$ I explain the distribution of both sentential and VP-adverbs. I propose that adverbs are predicates subcategorizing for their arguments at D-S. Like other predicates in English, adverbs are generated on the right of their subjects. Certain adverbs subcategorize for two arguments while other adverbs subcategorize for one argument. The selectional restrictions of the adverb are satisfied at S-S. Like other predicates, it is the maximal projection (AdvP) that governs its subject(s). Government is an m-command relationship. The AdvP can move to the left, either through substitution to an empty X$\sp\prime$ adjunction site or through adjunction to an XP. The maximal projection of the subject forms a barrier out of which the AdvP can not move. Maximal projections, except AgrP, are barriers. In English, the AdvP can not move if the adverb is subcategorized for by the verb. This thesis also compares the explanatory powers of my approach to recent syntactic approaches by Iatridou, Travis and Zagona. I also relate my findings to the semantic approaches by Jackendoff, Bellert and Rochette.

Dissertação (mestrado) - Universidade Federal de Santa Catarina, Centro de Comunicação e Expressão, Programa de Pós-Graduação em Letras/Inglês e Literatura Correspondente, Florianópolis, 2009
Made available in DSpace on 2012-10-24T11:46:26Z (GMT). No. of bitstreams: 1 275832.pdf: 563544 bytes, checksum: d357f9cc6381c6d382cbe522b3e08723 (MD5)
The main goal of this study is showing the differences between the way the objective language of journalism is expressed in Portuguese and English texts and also showing how translators deal with those differences when they do exist, and if translators really contemplate that. This was done by concentrating on the result of the analysis of the use of adverbs in a parallel corpus of Portuguese and English, original and translated texts while observing some rules which are in the Manuals of Style and Writing of Newspapers in Portuguese (Folha de São Paulo) and English (New York Times). For the sake of organization and selection of terms, the computer program WordSmithTools is used. The results of this study is supposed to help both translators and translation teachers when evaluating, correcting, editing, analyzing, studying or teaching about translated texts. The results also promote some questioning about the translating act and the journalistic genre in Brazil and in the US.
O principal objetivo deste estudo é mostrar como a linguagem direta do jornalismo é expressa de maneira diferente tanto em português como em inglês e também mostrar como os tradutores lidam com estas diferenças quando elas realmente aparecem, e se os tradutores contemplam verdadeiramente estas diferenças. Isto foi feito focalizando o resultado da analise do uso dos advérbios em um corpus paralelo de textos originais e traduzidos em português e inglês ao se observar algumas regras encontradas no Manual de Estilo de Jornais em português (Folha de São Paulo) e inglês (New York Times). Para efeito de organização e seleção de termos, o programa de computador WordSmithTools foi usado. Os resultados deste estudo poderão ajudar tanto tradutores como professores de tradução ao avaliar, corrigir, editar, analisar, estudar ou ensinar a respeito de textos traduzidos. Os resultados também poderão promover algum questionamento a respeito do ato tradutório e sobre o gênero jornalístico no Brasil e nos EUA.

The question of adverbs and the meaning of the present perfect across varieties of English is central to sociolinguistic variationist methodologies that have approached the study of the present perfect (Winford, 1993; Tagliamonte, 1997; van Herk, 2008, 2010; Davydova, 2010; Tagliamonte, 2013). This dissertation attempts to disentangle the effect of adverbial support from the three canonical readings of the present perfect (Resultative, Experiential and Continuative). Canadian English, an understudied variety of English, is used to situate the results seen in the Early Modern English data. Early Modern English reflects the time period in which English has acquired the full modern use of the present perfect with the three readings. In order to address both these questions and current controversies over statistical models in sociolinguistics, different statistical models are used: both the traditional Goldvarb X (Sankoff, Tagliamonte and Smith, 2005) and the newer mixed-effects logistic regression (Johnson, 2009). What is missing from the previous literature in sociolinguistics that advocates logistic mixed-effects models, and provided in this dissertation, is a clear statement of where they are inappropriate to use and their limitations. The rate of adverbial marking of the present perfect in Canadian English falls between rates reported for US and British English in previous studies. The data show in both time periods that while adverbs are highly favored in continuative contexts, they are strongly disfavored in experiential and resultative contexts. In Early Modern English, adverbial support functions statistically differently for resultatives and experientials, but that difference collapses in the Canadian English sample. Both this and the other linguistic contexts support a different analysis for each set of data with respect to adverbial independence from the meaning of the present perfect form. Finally, when the focus of the analysis is on linguistic rather than social factors, both the traditional and newer models provide similar results. Where there are differences, however, these can be accounted for by the number of tokens and different estimation techniques for each model.

The purpose of this study is to investigate the possible differences in the way American and Swedish journalists writing in English use maximizers, i.e. intensifying adverbs. In order to fulfill the purpose, information about how maximizers are used in two different sub-corpora, namely SWENC (The Swedish-English Corpus) and TIME (Time Corpus of American English) is collected. The data in SWENC has been collected from various websites where the crucial criterion was that the authors of the articles must have Swedish as their first language. The data from TIME has been collected from Time Magazine’s online corpora which is freely available on the Internet. The results show that there are some differences in the way Swedes and Americans use maximizers. The number of tokens for each maximizer does not differ to a great extent between the two corpora. However, there are larger deviances in the use of collocations and semantic prosody. The conclusion drawn from this is that there are indeed some differences in how American and Swedish journalists writing in English use maximizers, although they are not very many.

Frailty affects 10% of adults aged 65 years and older. It denotes loss of an individual’s body reserves, which increases vulnerability to developing adverse health outcomes such as death, disability, and institutionalization. Consequently, frailty has been described as the most problematic expression of ageing. Having good understanding of specific conditions influencing development of frailty and its effects holds the key to slowing its progression and mitigating its adverse outcomes. To this end, pathways to frailty and its adverse outcomes are the focus of my thesis. I begin with a literature review to assemble evidence on frailty pathways and instruments. Guided by this evidence and using the working framework of the Canadian Initiative on Frailty and Aging as the template, frailty pathways incorporating physical, psychological, and social conditions are conceptualized. Arguing that narrower physical frailty specifications are more suitable for investigating these pathways, I develop them based on the frailty phenotype. In my first two papers, I use data of 4,638 respondents aged 65 to 89 years from the English Longitudinal Study of Ageing to demonstrate construct, concurrent, and predictive validity of two physical frailty specifications. Adopting the specification with three indicators for latent growth curve analysis in my third paper, I show that chronic disease, allostatic load, low physical activity, cognitive impairment, depressive symptoms, poor social support, and poor social integration are predictors, mediators, or moderators on pathways to physical frailty. In my fourth paper, discrete time survival analysis reveals that low physical activity and cognitive impairment are mediators on pathways from physical frailty to death. In my fifth paper, autoregressive cross-lagged analyses demonstrate that these two conditions and depressive symptoms are mediators on pathways from physical frailty to activity limitation. All these conditions represent potentially modifiable targets for population-level interventions to address physical frailty in older people.

The use of routinely collected, or administrative, data for measuring and monitoring patient safety in primary care is a relatively new phenomenon. With increasing availability of data from different sources and care settings, their application for adverse event surveillance needs evaluation. In this thesis, I demonstrated that data routinely collected from primary care and secondary care can be applied for internal monitoring of adverse events at the general practice-level in England, but these data currently have limited use for safety benchmarking in primary care. To support this statement, multiple approaches were adopted. In the first part of the thesis, the nature and scope of patient safety issues in general practice were defined by evidence from a literature review and informal consultations with general practitioners (GPs). Secondly, using these two methods, measures of adverse events based on routinely collected healthcare data were identified. Thirdly, clinical consensus guided the selection of three candidate patient safety indicators for investigation; the safety issues explored in this thesis were recorded incidents with designated adverse event diagnostic codes and complications associated with two common diseases, emergency admissions for diabetic hyperglycaemic emergencies (diabetic ketoacidosis, DKA and hyperglycaemic hyperosmolar state, HHS) and cancer. In the second part of the thesis, the contributions of routinely collected data to new knowledge about potentially preventable adverse events in England were considered. Data from a primary care trust (NHS Brent), national primary care data (from the General Practice Research Database, GPRD) and secondary care data (Hospital Episode Statistics, HES) were used to explore the epidemiology of, and patient characteristics associated with, coded adverse events and emergency admissions for diabetic hyperglycaemic emergencies and cancer. Low rates of adverse events were found, with variation by individual patient factors. Finally, recommendations were made on extending the uses of routinely collected data for patient safety monitoring in general practice.

This study examines adverbial clause usage by women and men in three contexts: (1) single-gender and mixed-gender meetings of a US social club, (2) single-gender and mixed gender dyads speaking Ecuadorean Spanish, and (3) French speakers in informal interviews. The English results generally support past findings that women tend to use adverbial clauses to weaken their expressed commitments to the propositions they express, while men tend to use more of the clauses that strengthen the expressed commitments. However, the situation affects the specific clause types used, and the English results are not replicated in the Spanish or French data.

Le point de départ de cette recherche est le constat des difficultés persistantes rencontrées par les francophones de niveau intermédiaire à avancé lors de la production de textes en anglais, dans des contextes personnels ou professionnels. Les premiers outils utilisés pour remédier à ces erreurs, les correcteurs grammaticaux automatiques, ne prennent pas en compte de nombreuses erreurs produites par les francophones utilisant l'anglais, notamment car ces correcteurs sont rarement adaptés à un public ayant l'anglais comme L2. Nous proposons d'identifier précisément les difficultés rencontrées par ce public cible à partir du relevé des erreurs dans un corpus adapté, et d'élaborer une modélisation linguistique des erreurs et des corrections à apporter. Cette modélisation est fondée sur une analyse linguistique approfondie des phénomènes concernés, à partir d'indications grammaticales, d'études de corpus, et de l'analyse des segments erronés. La validité de l'utilisation de méthodes linguistiques est établie par l'implémentation informatique des règles de détection et de correction, suivie de l'évaluation des résultats de l'application de ces règles sur des corpus d'anglais L1 et L2
The starting point of this research is the observation that French speakers writing in English in personal or professional contexts still encounter grammatical difficulties, even at intermediate to advanced levels. The first tools they can reach for to correct those errors, automatic grammar checkers, do not offer corrections for a large number of the errors produced by French-speaking users of English, especially because those tools are rarely designed for L2 users. We propose to identify the difficulties encountered by these speakers through the detection of errors in a representative corpus, and to create a linguistic model of errors and corrections. The model is the result of the thorough linguistic analysis of the phenomena at stake, based on grammatical information available in reference grammars, corpus studies, and the analysis of erroneous segments. The validity of the use of linguistic methods is established through the implementation of detection and correction rules in a functional platform, followed by the evaluation of the results of the application of those rules on L1 and L2 English corpora

In light of recent surveys on gender equality and English proficiency in Sweden, this study examines the adjectives and adverbs which describe the women and men in two EFL-textbooks for English 6 in the upper secondary school, namely Blueprint: Version 2.0 B and WWE: World Wide English. Textbooks are often integral for language learning, and considering that students of Swedish upper secondary school are required to take English 6, the exposure of EFL-textbooks are significant. Thus, it becomes relevant to analyse how the language of EFL-textbooks depicts women and men. The actions, social roles and occupations of the female and male characters are analysed to find any discrepancies in how women and men are portrayed and whether these agree with the guidelines of the Swedish National Agency for Education. These guidelines require teaching to be carried out with consideration to fundamental democratic rights and should strive to promote equality between groups. In addition, the study builds upon previous research within the field of linguistics as well as social sciences carried out by Fairclough, Foucault, Lucy, Butler and others which concerns discourse, linguistic relativism, gender theory etc. The hypothesis of this study is that there is still a discrepancy in how women and men are depicted in EFL-textbooks. By employing a mixed method approach which includes quantitative data and statistics and qualitative discourse analysis which highlights indications of unequal description of gender, it can be concluded that women and men are described differently, and often in terms of dichotomies, with adjectives, adverbs and the social roles that they are assigned. These descriptions may consequently result in that students that use these textbooks as part of their learning process may adopt these values. Thus, some descriptions violates the goals and guidelines for gender equality, prescribed by the Swedish National Agency for Education (Skolverket).

The literature on J. K. Rowling’s Harry Potter is prolific. People have written on various topics dealing with issues such as translation, etymology and diverse areas concerning the language. In this essay, I examine whether linguistic features such as reporting verbs, adverbs of manner and adjectives contribute to the depiction of heroic and villainous characters. Before conducting this research, my assumptions were that there would be a great difference in the value of the words depending on the character they were associated with. I wanted to see if the heroic characters used verbs and adverbs with positive connotations, and the villainous characters verbs and adverbs with negative connotations. I also wanted to know if the adjectives describing the characters could, in themselves, clearly indicate whether a character was a hero or a villain.

The results of my research suggested that the choice of particular verbs and adverbs contributed only indirectly to the depiction of the characters. Without context, it was not possible to know if the character was a hero or a villain simply by identifying the verbs and adverbs used to describe their speech. By contrast, the choice of particular adjectives did appear to indicate more clearly whether a character was hero or villain. Finally, the results of my research indicated that context, rather than the use of particular linguistic features was often the most important factor in contributing to the portrayal of characters in the novel.

Cette étude vise à fournir une description détaillée de l’évidentialité en anglais dans une approche contrastive avec le tibétain. Elle repose sur un corpus spécialement recueilli au Tibet et en Angleterre (TSC et CSC/LAC, 2010-2012, 10 h.). Le tibétain possède un système évidentiel complexe et grammaticalisé dont la description peut fournir une grille d’analyse préliminaire pour notre étude de l’évidentialité en anglais. Des exemples authentiques et des données quantitatives issus du corpus nous permettent d’illustrer et de compléter les descriptions du groupe verbal tibétain dans plusieurs travaux pionniers antérieurs à la présente recherche (Tournadre & Sangda Dorje 1998) afin de poser les bases de la sémantique évidentielle. Les marqueurs évidentiels qui émergent dans les parties tibétaine et anglaise du corpus sont examinés afin de déterminer les paramètres qui motivent leur utilisation. Les marqueurs tibétains sont principalement grammaticaux et intégrés à des paradigmes syntaxiques : copules, suffixes verbaux et enclitiques. Les marqueurs évidentiels anglais sont lexicaux ou semi-grammaticaux : verbes de perception, verbes de cognition, verbes de discours, modaux, adverbes, conjoints, parenthétiques et marqueurs de discours (Nuyts 2001a, Cappelli 2007, Sanders & Sweetser 2009, Mortensen 2010, Whitt 2010, Gisborne 2010, Miller 2008, Boulonnais 2010, Gurajek 2010, Kaltenböck et coll. 2011, Heine 2013). Cette description de l’évidentialité en tibétain et en anglais nous permet d’analyser les conséquences d’un rendu grammatical ou lexical de cette notion (Talmy 2000, Bybee et coll. 1994, Nuyts 2001a, Boye & Harder 2009). Des analyses quantitatives et qualitatives du corpus contrastif et sur large corpus illustrent les différences de complexité, d'optionalité, de fréquence, de restriction sémantique, de prise en charge, de statut informatif, et de stratégie de discours dans les deux systèmes.Enfin, nous réévaluons la dichotomie, souvent considérée comme acquise, entre évidentialité grammaticale et lexicale (Aikhenvald 2004). Nous avançons que les marqueurs évidentiels des deux langues se positionnent à différents endroits sur un continuum multidimensionnel lexique-grammaire. Cette étude ne remet pas en question que le système évidentiel tibétain est plus grammaticalisé que le système anglais, mais démontre que ce dernier présente tous les signes d’une grammaticalisation partielle. Nous concluons ainsi que l’évidentialité est une notion pertinente et nécessaire pour une description linguistique complète de l’anglais
This study aims to give a detailed description of evidentiality in English in contrast with Tibetan. It is based on a specialised corpus collected in Tibet and in England (TSC and CSC/LAC, 2010-2012, 10 h.).Tibetan has a complex and grammaticalised evidential system, and its description can provide a preliminary analytical grid for a semantic assessment of English evidentiality. Athentic examples and quantitative data from the corpus illustrate and supplement the analyses of the Tibetan verb phrase from previous research (Tournadre & Sangda Dorje 1998), in order to lay the foundation of the semantics of evidentiality. The evidential markers that emerge in the Tibetan and English sections of the corpus are examined so as to determine the parameters that motivate their usage. Tibetan evidentials are mainly grammatical and paradigmatised: copulas, verb suffixes and enclitics. English evidentials are either lexical or semi-grammatical: perception verbs, cognition verbs, speech verbs, modals, adverbs, conjuncts, parentheticals and discourse markers (Nuyts 2001a, Cappelli 2007, Sanders & Sweetser 2009, Mortensen 2010, Whitt 2010, Gisborne 2010, Miller 2008, Boulonnais 2010, Gurajek 2010, Kaltenböck et al. 2011, Heine 2013).This survey of Tibetan and English evidentiality provides precise data for the analysis of the consequences of a grammatical or a lexical rendering of this notion (Talmy 2000, Bybee et al. 1994, Nuyts 2001a, Boye & Harder 2009). Qualitative and quantitative evidence illustrates the differences in complexity, optionality, frequency, semantic restriction, speaker commitment, informative status and discourse strategy in the two systems. Finally, this study reassesses the assumed dichotomy between grammatical and lexical evidentiality (Aikhenvald 2004), arguing that evidentials in both Tibetan and English inhabit a lexicon-grammar multidimensional continuum, merely in different positions. This study does not question that the Tibetan evidential system is more grammaticalised than the English one, but it shows that the latter presents all the signs of partial grammaticalisation. This reassessment leads to the conclusion that evidentiality is a relevant and necessary notion for a thorough linguistic description of English

碩士
國立嘉義大學
外國語言學系研究所
106
Linguistics, especially syntax, is usually one of the most difficult subjects for college students. In high school, students learned sentence pattern but only little, or none, about the logic behind the sentence forming, so some of students were confused by those principles and parameters and believed learning syntax could only mess up the idea they had already known. Chou & Bartz (2007) suggested that over 50% of student’s syntactic errors are from L1 transfer and provide strong evidence to support the teaching efficiency of non-native English-speaking teachers (NNESTs). In that case, to understand the logic of sentence forming is extremely important for NNESTs. However, for the first data we collected, there were 60.5% about the misplacing of adjuncts and complements among all errors, and 27.7% for the second data. It shows that although linguistics learners have basic knowledge about the basic sentence forming, it is difficult for them to tell the difference between adjuncts and complements. Telling the difference between adjuncts with complements is very important due to complements are arguments but adjunct are not (Haegeman 1994). Besides adjuncts and complements, landing site of wh-movement is also one of the big problems for learners. In addition, we also find out that students tend to mark ‘adverb of time’ and ‘adverb of place’, known as bare-NP (Larson 1985), as a noun or preposition due to their function. Linguistics teachers should notice this phenomenon and pay more attention on X-bar theory, so the linguistics learners may acquire the idea more clearly. We believe that second language acquisition, linguistics learning and language teaching is a circle. We firstly acquire a second language; then go deeper to understand the process of language, and finally use the knowledge we learned from linguistics to provide feedback to the other learners. The aim of this study is to identify the gaps between them for linguistics and language teachers to find a better way of teaching.

Linguistic research on adverbs has taken many forms: typological, morphological, syntactic, semantic and pragmatic. However, little work has been conducted on adverbs using the tools of quantitative sociolinguistics, and most of that work has focused solely on morphological variation of the -ly suffix. This work addresses the lacuna by examining two adverb phenomena using quantitative variationist methodology. Data come from two large, socially stratified, sociolinguistic corpora of vernacular English. The two corpora contain data collected in Ontario, Canada and in Northern England, and are comprised of the speech of over 150 speakers across all age groups. In the first case study, I examine a claim in usage guides (e.g., Swan 2001) that North American English widely permits pre-auxiliary adverbs in canonical, declarative sentences, while British English prohibits them unless accompanied by contrastive stress. As I show, the varietal differences in speech are not only minimal and unrelated to stress, but instead are highly circumscribed. In addition, I demonstrate that the positioning of adverbs observed here must involve post-syntactic processes. The second case study examines variability in the discourse adverb "actually" and several related adverbials (e.g., "really" and "in fact") and examines the path of grammaticalization (Traugott & Dasher 2002) in the two communities. I demonstrate that Canadians, regardless of sex or education level, prefer the more grammaticalized forms of "actually"; in the UK, the more grammaticalized use is less common, though some young men are leading a shift to the more grammaticalized pattern.

Dissertação de mestrado, Processamento de linguagem natural e tecnologia da linguagem humana, Faculdade de Ciências Humanas e Sociais, Universidade do Algarve. UFR Sciences du Langage, de L’Homme et de la Societé, Université de Franche-Comté, 2012
Esta dissertação tem por objetivo a investigação do padrão colocacional formado por verbo e advérbio terminado em –mente em português em vista de sua extração de corpora e sua tradução automática para o inglês. O trabalho envolve o processamento computacional de um corpus do português; o desenvolvimento de um conjunto de regras que permitam um melhor processamento desse padrão, sobretudo resolvendo o problema de coordenação adverbial; um teste da intuição de falantes nativos do português em vista da identificação do valor colocacional do padrão linguístico estudado; uma avaliação da sensibilidade de medidas de associação para a identificação de colocações com este padrão; o desenvolvimento de um classificador automático de colocações com base em métodos de aprendizagem supervisionada; a construção de um léxico bilíngue deste tipo de colocações; e a avaliação da tradução automática deste padrão para o inglês. Na primeira fase do estudo, um corpus do português de grande porte, o CETEMPúlico, composto por 191 milhões de palavras de textos jornalísticos, foi processado computacionalmente por meio da cadeia de processamento STRING, que faz desde a segmentação do texto até sua análise sintática. Nesta fase, uma série de regras com vistas a um melhor processamento de casos de coordenação adverbial em português foram criadas e incorporadas na STRING. Os resultados obtidos para desambiguação de partes do discurso consistem em uma medida-f de 0.724, já para chunking e extração de dependências, uma medida-f de 0.810 foi obtida. Uma vez processado o corpus, 65.535 dependências sintáticas entre verbo e advérbio terminado em –mente foram extraídas. Em seguida, uma série de filtros foram aplicados ao resultado da extração para que fossem excluídos desde o início casos que não apresentavam potencial para formar colocações. Primeiramente, um filtro de frequência que excluía pares que ocorrem menos de 5 vezes no corpus foi adotado. Também foram excluídos bigramas que incluíam verbos de ligação, assim como bigramas que incluíam classes adverbiais que apresentam pouco ou nenhum potencial colocacional. Uma classificação previamente existente de advérbios terminados em –mente em português foi utilizada para este fim. Esta classificação foi estendida em aproximadamente 500 advérbios e em seguida incorporada na cadeia de processamento STRING como parte do presente estudo. Uma série de critérios propostos para a classificação de advérbios terminados em –ment, em francês, foi tomada como o conjunto de princípios linguísticos que serviram de base para a classificação dos advérbios em português. Após a fase de filtragem, 5.793 pares de verbo e advérbio terminado em –mente restaram da extração. Para que se chegasse a uma lista de colocações deste padrão em português, esses 5.793 pares, considerados como o conjunto de pares-candidatos, passaram por uma classificação manual que etiquetava os pares como “colocação” ou como “não colocação”. Uma série de testes linguísticos foram desenvolvidos para a classificação dos pares. O objetivo desses testes era facilitar a identificação deste tipo de colocação por meio de princípios sintático-semânticos que discutivelmente refletem a existência de um caráter colocacional em um par ou grupo de palavras. Como resultado da classificação manual, 501 bigramas foram considerados colocações dos 5.793 candidatos. Pôde-se notar que a frequência dos pares no corpus estava de certa forma ligada ao seu caráter colocacional, uma vez que 60 por cento dos pares mais frequentes, contra 8.6 por cento do total de candidatos, foram considerados como casos de colocação. Para averiguar a intuição de falantes nativos do português a respeito desse padrão colocacional, uma tarefa de classificação foi desempenhada com uma amostra de 30 pares selecionados aleatoriamente da lista de candidatos – 15 tendo sido previamente classificados como colocações, e 15 como não colocações. Vinte e um falantes nativos do português foram recrutados para a tarefa de classificação, dos quais 13 eram falantes nativos do português europeu, e 8 do português brasileiro. Foi possível concluir com o resultado dessa experiência que o padrão colocacional tratado é extremamente problemático no que diz respeito a sua identificação. A medida Kappa de acordo entre anotadores para a amostra de 30 pares foi de 0.06, o que, embora possa ser interpretado como “leve acordo”, é ainda discutivelmente um valor consideravelmente baixo. A dificuldade de se explicar o próprio conceito de colocação assim como o tamanho reduzido da amostra seriam algumas das razões para o baixo nível de convergência alcançado. Haja vista a baixa qualidade dos resultados alcançados com a tarefa de classificação envolvendo falantes nativos do português, uma série de medidas de associação foram testadas em vista do padrão colocacional tratado. Primeiramente, constatou-se que o limiar de referência existente para a análise das medidas “t test” e “chi-quadrado” não apresenta resultados satisfatórios na identificação do tipo de colocação tratado. Em seguida, a sensibilidade dessas mesmas medidas, e também de “Informação Mútua”, “Log-Likelihood Ratio”, “Coeficiente Dice”, e “Unigram Subtuples”, foi testada com base em sua correlação com a classificação manual dos pares-candidatos. Constatou-se que “Informação Mútua”, “Log-Likelihood Ratio”, e “Unigram Subtuples” são as medidas de associação com maior correlação com a classificação manual, o que representa um desempenho satisfatório dessas medidas para a identificação do padrão colocacional sob estudo. Em seguida, técnicas de aprendizagem de máquina supervisionada foram utilizadas para que, a partir do conjunto de pares-candidatos classificados manualmente e seus respectivos valores de medidas associação, fosse possível treinar um classificador automático de colocações. Os resultados alcançados com esta experiência são extremamente promissores. O desempenho de quarenta e cinco classificadores disponíveis na ferramenta de aprendizagem de máquina WEKA foi testado com base em validação cruzada. O classificador que apresentou o melhor resultado foi “RotationForest”, que alcançou uma medida-f de 0.816 em um corpus de treino balanceado composto pelos 501 bigramas classificados como colocação, mais outros 501 bigramas classificados como não colocação. A estratégia que consiste em combinar diferentes classificadores por meio do algoritmo “Vote”, disponível na ferramenta WEKA, provou ser capaz de melhorar ainda mais os resultados. O desempenho de uma série de combinações foi testado, e o melhor resultado foi alcançado com a combinação “Rotation Forest” e “LMT”. Para validar os resultados obtidos, o classificador proveniente da combinação desses dois algoritmos foi testado em um corpus não visto, o NILC/São Carlos, consideravelmente menor que o corpus de treino. Considerando os casos de colocação que ocorrem nos dois corpora e excluindo-se casos de hápax legomena no NILC/São Carlos, o classificador alcançou uma medida-f de 0.733 para o corpus não visto, o que pode considerado bastante promissor devido à considerável diferença de tamanho entre os dois corpora. Após os testes com os diferentes métodos mencionados acima, compilou-se um léxico bilíngue português-inglês contendo o padrão colocacional tratado. Três corpora paralelos e um dicionário de colocações foram utilizados como fontes de referência para que versões equivalentes das colocações em inglês fossem estabelecidas. O dicionário adotado, o Oxoford Collocations Dictionary, foi considerado como fonte principal já que, mais que apenas ocorrências em um corpus paralelo, entradas em um dicionário de colocações atestam o verdadeiro valor colocacional das combinações em inglês. Uma vez construído o léxico, as equivalências deste tipo de colocação entre português e inglês foram utilizadas como referencia para a avaliação de três sistemas de tradução automática disponíveis gratuitamente na rede: Google Translate, Systranet, e Reverso. Exemplos do contexto de ocorrência dos pares em português foram extraídos do corpus CETEMPúblico e então traduzidos automaticamente para o inglês com esses três sistemas. Foi constatado que a tradução da maioria dos pares é correta no sentido de não infringir regras gramaticais da língua, mas, em contrapartida, a tradução sugerida para a maioria dos pares não reflete uma escolha lexical fluente em inglês. A avaliação da fluência das traduções foi feita tomando-se como referência medidas de associação calculadas para os pares com base em dados de frequência do corpus do inglês Collins Wordbanks. De modo geral, os resultados obtidos com este trabalho demonstram que o padrão linguístico formado por verbo e advérbio terminado em –mente impõe uma série de obstáculos a diversos níveis de processamento de linguagem natural, desde desambiguação de partes do discurso até tradução automática. A identificação do valor colocacional deste padrão também mostrou-se problemática, sobretudo quando a classificação de diversos anotares, ainda que falantes nativos do português, é considerada. Por fim, espera-se que os métodos testados no decorrer desta pesquisa possam não somente servir a um melhor tratamento computacional do padrão estudado em português, mas que possam também ser replicados a outros problemas linguísticos, sobretudo àqueles relacionados a termos compostos e expressões multipalavra em geral.

碩士
國立屏東教育大學
華語文教學碩士學位學程
101
The frequency with which people use the Chinese adverb of degree “h!n” in daily life is relatively high; however its contextual meanings differ according to how it appears with the adverb of negation, “bù,” in different character sequences. Elementary Mandarin textbook materials provide limited information on how to combine “h!n” and “bù.” Mandarin Chinese learners who are native English speakers are likely to utilize these two Chinese adverb components as “bù h!n” since that order correlates with English grammar, "not very". Nevertheless, the structures “h!n bù” and “bù h!n” indicate different emphasis. Hence, learners may not be expressing the exact meaning they intend. For the purpose of teaching these expressions to learners of Mandarin Chinese, this study explores its meanings and structures according to how native Mandarin speakers actually utilize “h!n bù + adjective” and “bù h!n + adjective.” The research draws data from the “Academia Sinica Balanced Corpus of Modern Chinese.” It also includes a survey study consisting of two questionnaires and follow-on interviews with Mandarin Chinese learners who are native English speakers. One questionnaire investigates how the participants perceive language use in daily life. The second questionnaire focuses on the participants’ syntactical concept of the two patterns. The questionnaires include some oral questions relative to the participants’ responses to the questionnaires. After integrating the corpus data and the data from the questionnaires, the research concludes with suggestions for teachers of Mandarin Chinese. The results show that the people (Mandarin speakers) around the Mandarin learners (native English speakers) greatly influence the latter's use of “h!n”. The learners recall Mandarin speakers more often using “h!n bù + adj.”; however, they also understand the two patterns include different meanings according to the syntax. Moreover, native English speakers learning Mandarin are unable to define which adjectives properly suit which pattern; consequently, they make many errors, judging from the standpoint of Mandarin language scholarship on this point. The results also show that learners of Mandarin rely on their native language when they are unable to distinguish the meanings of sentences in Mandarin. In other words, when they encounter difficulty trying to interpret a sentence because of its word order, they substitute the adverbial degree words or use another pattern of expression. This study suggests that teaching Mandarin should include a review of the adjectives that accompany negative prefixes, a classification of similar adverbs of degree, and explanations and instructions about how the use of “h!n” and “bù” differs semantically and colloquially. In addition, it is important to remind the learners how to express certain ideas euphemistically. Mandarin teachers need to reflect on how they learned a foreign language and recall the difficulties of learning the idiosyncrasies of that language. Finally, although the structure of some syntax may be abstruse, it is important for Mandarin teachers not to let this issue interfere with their students' confidence with learning Mandarin.