Robust handling of out-of-vocabulary words in deep language processing

Tese de doutoramento, Informática (Ciências da Computação), Universidade de Lisboa, Faculdade de Ciências, 2014

Para um modelo ontogenético da escrita infantil no 1.º ciclo de ensino básico : a emergência do período como unidade de organização textual

Tese de doutoramento, Linguística (Linguística Aplicada), Universidade de Lisboa, Faculdade de Letras, 2015

Métodos probabilísticos e estatísticos na gestão de fogos florestais

Tese de doutoramento, Estatística e Investigação Operacional (Probabilidades e Estatística), Universidade de Lisboa, Faculdade de Ciências, 2014

Da criatividade linguística à tradução : uma abordagem das unidades polilexicais em Mia Couto

Tese de doutoramento, Linguística (Linguística Aplicada), Universidade de Lisboa, Faculdade de Letras, 2015

Ensino da música em Portugal (1868-1930) : uma história da pedagogia e do imaginário musical

Tese de doutoramento, Educação (História da Educação), Universidade de Lisboa, Instituto de Educação, 2015

Wildfire susceptibility modelling in mainland Portugal

Tese de doutoramento, Geografia (Geografia Física), Universidade de Lisboa, Instituto de Geografia e Ordenamento do Território, 2015

Avaliação dos níveis de ansiedade e depressão e sua correlação com a intensidade de dor em doentes oncológicos com doença avançada e progressiva

Tese de mestrado, Cuidados Paliativos, Faculdade de Medicina, Universidade de Lisboa, 2015

Cancer-associated fibroblasts in brain metastases

Tese de mestrado. Oncobiologia, Faculdade de Medicina, Universidade de Lisboa, 2015

A study of Aldred's multiple glosses to the Lindisfarne Gospels

Aldred, the glossator of the Lindisfarne Gospels, presents himself as carefully rendering the Latin lemmata in front of him, in terms of both their internal structure and meaning. His work includes a very high number of multiple glosses, which often attempt to clarify the polysemous character of a lemma or to provide additional information. This paper explores the multiple glosses including different lexemes which Aldred added to lexical lemmata in Mark’s Gospel in an attempt to establish whether there is any correlation between Aldred’s ordering practices and the frequency with which he used the interpretamenta to render those lemmata. The results of the study show some preference for placing the interpretamentum which most commonly renders the Latin lemma in first position, although Aldred’s practice is not fully consistent.

Cappadocia: where Greek prepositions come to die

The Cappadocian variety of Ulaghátsh is unique among the Greek-speaking world in having lost the inherited preposition ‘se’. The innovation is found with both locative and allative uses and has af-ected both syntactic contexts in which ‘se’ was originally found, that is, as a simple preposition (1) and as the left-occurring member of circumpositions of the type ‘se’ + NP + spatial adverb (2). (1) a. tránse ci [to meidán] en ávʝa see.PST.3SG COMP ART.DEF.SG.ACC yard.SG.ACC COP.3 game.PL.NOM ‘he saw that in the yard is some game’ (Dawkins 1916: 348) b. ta erʝó da qardáʃa évɣan [to qonáq] ART.DEF.PL.NOM two ART.DEF.PL.NOM friend.PL.NOM ascend.PST.3PL ART.DEF.SG.ACC house.SG.ACC ‘the two friends went up to the house’ (Dawkins 1916: 354) (2) émi [ta qonáca mésa], kiríʃde [to ʝasdɯ́q píso] enter.PST.3SG ART.DEF.PL.NOM house.PL.ACC inside hide.PST.3SG. ART.DEF.SG.ACC cushion.SG.ACC behind ‘he went into the houses and hid behind the cushions’ (Dawkins 1916: 348) In this paper, we set out to provide (a) a diachronic account of the loss of ‘se’ in Asia Minor Greek, and (b) a synchronic analysis of its ramifications for the encoding of the semantic and grammatical functions it had prior to its loss. The diachronic development of ‘se’ is traced by comparing the Ulaghátsh data with those obtained from Cappadocian varieties that have neither lost it nor do they show signs of losing it and, crucially, also from varieties in which ‘se’ is in the process of being lost. The comparative analysis shows that the loss first became manifest in circumpositions in which ‘se’ was preposed to the complement to which in turn a wide range of adverbs expressing topological relations were postposed (émi sa qonáca mésa > émi ta qonáca mésa). This finding is accounted for in terms of Sinha and Kuteva’s (1995) distributed spatial semantics framework, which accepts that the elements involved in the constructions under investigation—the verb (émi), ‘se’ and the spatial adverb (mésa)—all contribute to the expression of the spatial relational meaning but with differences in weighting. Of the three, ‘eis’ made the most minimal contribution, the bulk of it being distributed over the verb and the adverb. This allowed for it to be optionally dropped from circumpositions, a stage attested in Phlo-tá Cappadocian and Silliot, and to be later completely abandoned, originally in allative and subsequently in locative contexts (earlier: évɣan so qonáq > évɣan to qonáq; later: so meidán en ávʝa > to meidán en ávʝa). The earlier loss in allative contexts is also dealt with in distributed semantics terms as verbs of motion such as έβγαν are semantically more loaded than vacuous verbs like the copula and therefore the preposition could be left out in the former context more easily than in the latter. The analysis also addresses the possibility that the loss of ‘se’ may ultimately originate in substandard forms of Medieval Greek, which according to Tachibana (1994) displayed SPATIAL ADVERB + NP constructions. Applying the semantic map model (Croft 2003, Haspelmath 2003), the synchronic analysis of the varieties that retain ‘se’ reveals that—like many other allative markers crosslinguistically—it displays a pattern of multifunctionality in expressing nine different functions (among others allative, locative, recipient, addressee, experiencer), which can be mapped against four domains, viz. the spatiotemporal, the social, the mental and the logicotextual (cf. Rice & Kabata 2007). In Ulaghátsh Cappadocian, none of these functions is overtly marked as such. In cases like (1), the intended spatial relational meaning is arrived at through the combination of the syntax and the inherent semantics of the verb and the zero-marked NP as well as from the context. In environments of the type exemplified by (2), the adverb contributes further to the correct interpretation. The analysis additionally shows that, despite the loss of ‘se’, Ulaghátsh patterns with all other Cappadocian varieties in one important aspect: Goal and Location are expressed similarly (by zero in Ulaghátsh, by ‘se’ in the other varieties) whereas Source is being kept distinct (expressed by ‘apó’ in all varieties). Goal-Location polysemy is very common across the world’s languages and, most crucially, prevails over other possible polysemies in the tripartite distinction Source—Location—Goal (Lestrade 2010, Nikitina 2009). Taking into account this empirical observation, our findings suggest that the reor-anisation of spatial systems can have a local effect—in our case the loss of a member of the prepositional paradigm—but will keep the original global picture intact, thus conforming to crosslinguistically robust tendencies.  References Croft, W. 2001. Radical Construction Grammar: Syntactic Theory in Typological Perspective. Oxford: Oxford University Press. Dawkins, R. M. 1916. Modern Greek in Asia Minor: A Study of the Dialects of Sílli, Cappadocia and Phárasa with Grammar, Texts, Translations and Glossary. Cambridge: Cambridge University Press. Haspelmath, M. 2003. The geometry of grammatical meaning: semantic maps and cross-linguistic comparison. In M. Tomasello (Ed.), The New Psychology of Language, Volume 2. New York: Erlbaum, 211–243. Lestrade, S. 2010. The Space of Case. Doctoral dissertation. Radboud University Nijmegen. Nikitina, T. 2009. Subcategorization pattern and lexical meaning of motion verbs: a study of the source/goal ambiguity. Linguistics 47, 1113–1141. Rice, S. & K. Kabata. 2007. Cross-linguistic grammaticalization patterns of the allative. Linguistic Typology 11, 451–514. Sinha, C. & T. Kuteva. 1995. Distributed spatial semantics. Nordic Journal of Linguistics 18:2, 167–199. Tachibana, T. 1994. Syntactic structure of spatial expressions in the “Late Byzantine Prose Alexander Romance”. Propylaia 6, 35–51.

Codeswitching in Igbo-English Bilingualism

This study evaluates the Matrix Language Frame model of codeswitching with Igbo-English data and concludes that the data can indeed be considered a classic case of codeswitching, in that a Matrix Language can be clearly identified in bilingual clauses. It establishes this through both qualitative and quantitative analyses that make use of the typological contrasts between Igbo and English to uncover supportive evidence for the Matrix Language Frame model and its associated three principles: the Matrix Language Principle, the Asymmetry Principle, and the Uniform Structure Principle. The investigation goes one step further by using spectrograms and the analysis of vowel harmony between English free morphemes and Igbo bound affixes to demonstrate that two phonologies can co-exist in codeswitching and that codeswitching forms are essentially pronounced with a phonology that does not entirely resemble that of the Matrix Language variety. Furthermore, the study finds that the same language production mechanisms as detailed under the Matrix Language Frame model and its associated three principles underlie both single word and multi-word codeswitching. That is, the present study, like those before it adopting the Matrix Language framework (see Amuzu 2010: 277), underlines the importance of the assumptions underpinning the Matrix Language Principle: (1) that language production is modular; (2) that lexical structure is both complex and abstract; and (3) that languages in contact divide responsibilities in what they may contribute toward lexical structure during the production of mixed constituents. Moreover, the study finds that Igbo-English bilinguals can always sustain ready access to their mother tongue mental lexicon during online speech production and thus Igbo-English may duly be described as a ‘classic’ case of codeswitching.

A Bilingual Thesaurus of Everyday Life in Medieval England: Some Issues at the Interface of Semantics and Lexicography

This paper reports on issues at the interface between semantics and lexicography that arose out of the data collection and classification of vocabulary in Anglo-Norman and Middle English in order to create a bilingual thesaurus of everyday life in medieval England. The Bilingual Thesaurus project is based at Birmingham City University and the University of Westminster. Issues to be resolved included the definition of an occupational domain; the creation of a methodology of data collection; the delimitation of domain-specific vocabulary; making distinctions between sense and usage; and the categorisation of the lexical items. Some of these issues are general to thesaurus-making, some are specific to the making of historical thesauruses, while some are unique to the production of a thesaurus of two languages whose use overlapped for several centuries in the late medieval period in England.

Technical Vocabulary and Medieval Text Types: A Semantic Field Approach

This paper seeks to discover in what sense we can classify vocabulary items as technical terms in the later medieval period. In order to arrive at a principled categorization of technicality, distribution is taken as a diagnostic factor: vocabulary shared across the widest range of text types may be assumed to be both prototypical for the semantic field, but also the most general and therefore least technical terms since lexical items derive at least part of their meaning from context, a wider range of contexts implying a wider range of senses. A further way of addressing the question of technicality is tested through the classification of the lexis into semantic hierarchies: in the terms of componential analysis, having more components of meaning puts a term lower in the semantic hierarchy and flags it as having a greater specificity of sense, and thus as more technical. The various text types are interrogated through comparison of the number of levels in their hierarchies and number of lexical items at each level within the hierarchies. Focusing on the vocabulary of a single semantic field, DRESS AND TEXTILES, this paper investigates how four medieval text types (wills, sumptuary laws, petitions, and romances) employ technical terminology in the establishment of the conventions of their genres.

A divulgação de informação sobre sistemas integrados de gestão (SIG), sustentabilidade empresarial (SE) & responsabilidade social empresarial (RSE), nos websites das empresas portuguesas certificadas "Uma análise exploratória"

A Internet é um meio priviligiado de comunicação, no qual a informação não conhece fronteiras, e a sua divulgação toma uma dimensão universal. No mundo empresarial a comunicação é hoje tida, como o elemneto fulcral dos processos de negócio , em que interação com as partes interessadas, pode ser traduzida em termos do seu exposente máximo pelo website da empresa. Muita da informação hoje divulgada à sociedade pela empresa, visa o seu próprio escrutínio público, na strês dimensões (económica, ambiental e social) do Desenvolvimento Sustentável (DS). A presente investigação visou uma análise exploratória dos websites de 523 empresas que se encontravam certificadas, no fianl do ano de 2011, no âmbito dos sistemas de gestão da qualidade (ISO 9001), ambiente (ISO 14001), segurança e saúde do trabalho (OHSAS 18001). O obketo de estudo proposto visou conhecer quais os conteúdos que são frequentemente mais divulgados, e que perfil tem a empresa portuguesa de Gestão (SIG), Sustentabilidade Empresarial (SE) e Responsabilidade Social Empresarial (RSE). O método de investigação baseou-se na técnica da Análise de Conteúdo, a qual permitiu quantificar por categoria de análise (indicador) e subcategorias de análise (itens), os conteúdos da informação divulgada no website da empresa. A visão holística sobre a dimensão dos contéudos disponibilizados no website da empresa, foir determinada (medida) com auxílio de divulgação no website das empresas, são: a Missão, a Política da Qualidade, Ambiente e Segurança, o Código de Ética & Conduta Empresarial, o Relatório de Sustentabilidade, o Relatório e Contas, os Indicadores & Índices Financeiros, os certificados e marcas do Sistema de Gestão da Qualidade (SGQ), do trabalho (SGSST) e, por fim, os Projetos de Envolvimento com a Comunidade no âmbito da RSE. Em termos de perfil, as empresas portugueas de maior dimensão (volume de vendas), e as sociedades anónimas (S.A:), divulgam mais informação no website, que as outras. Não tendo sido porém, encontrada evidência estatística significativa no estudo da relação das empresas localizadas no distrito de Lisboa & Setúbal, das empresas do setor de atividade secundário (2.º), e das empresas com exposição aos media (rankings nacionais), com a maior divulgação de informação no website da empresa.

A Divulgação de Informação sobre o Capital Humano nas empresas cotadas em Portugal. Uma análise empírica de 2008 a 2012.

Esta investigação empírica tem como objetivo principal analisar o nível de divulgação de informação voluntária sobre o capital humano das empresas cotadas na Euronext Lisbon de 2008 a 2012, bem como identificar os seus fatores determinantes. Estudamos os relatórios anuais de contas consolidadas recorrendo à técnica de análise de conteúdo, numa amostra composta por 46 empresas. Para determinar o nível de divulgação de informação sobre o capital humano desenvolvemos um índice de divulgação (IDVCH), composto por 45 indicadores, baseado no estudo de Silva et al. (2012). Os dados recolhidos para a construção do índice foram codificados de forma binária: 0 para “não divulga” e 1 para “divulga”. Os resultados obtidos apontam para um nível médio de divulgação de informação sobre o capital humano de 0,46, verificando-se uma evolução positiva e estatisticamente significativa do IDVCH durante os cincos anos. Os setores de atividade que apresentam maiores índices de divulgação são o setor financeiro, o setor de telecomunicações e o setor público. Identificámos através da análise univariada e multivariada que os fatores que influenciam positivamente o nível de divulgação de informação sobre o capital humano são: a rendibilidade, a dimensão do conselho de administração, a dispersão geográfica dos colaboradores, e a integração da sociedade no índice PSI-20. Não se obteve significância estatística para se concluir sobre os resultados das variáveis independentes: a dimensão da organização, o endividamento, o tipo de auditor, o setor de atividade e a idade da organização. O modelo de regressão linear múltipla, (em que considerámos o IDVCH como variável dependente), é explicado na ordem dos 40% pelas variáveis independentes que analisámos. A originalidade desta dissertação incide no facto de tratarmos isoladamente a componente do capital humano, e de apurarmos quais os fatores que influenciam o seu nível de divulgação nas empresas cotadas em Portugal.