Costa, Maria ArmandaIgnazio Coco, MorenoCarvalho, Mário Alexandre Gomes2019-03-202019-03-202019-01-242018-11-02http://hdl.handle.net/10451/37613Esta dissertação investiga a produção de linguagem em contexto visualmente situado, com um foco na produção de expressões referenciais e uma abordagem multi linguística que compara dados do Português Europeu com dados do Inglês. Literatura prévia na área da produção de linguagem visualmente situada tem estudado a produção de expressões referenciais com base na ideia de acessibilidade referencial (e.g. Ariel, 2014; Gundel, Hedberg, & Zacharski, 1993): formas referenciais mais reduzidas ou mais descritivas refletem referentes que são mais ou menos cognitivamente acessíveis. Com base neste ideia, houve uma procura dos fatores que informam a acessibilidade dos referentes. Foco-me nesta tese em três fatores previamente investigados pela literatura: a animacidade do referente, o nível de visual clutter e a presença de competição referencial. Foram encontrados efeitos para todos estes três fatores na produção de expressões referenciais (Arnold & Griffin, 2007; Clarke, Elsner, & Rohde, 2013; Fukumura & Gompel, 2011; Koolen, Krahmer, & Swerts, 2015), mas de uma forma isolada e testando diferentes aspetos das expressões referenciais. Através de uma experiência de produção de linguagem guiada, testo a animacidade referencial, o nível de 'visual clutter' (ambas como variáveis independentes) e a presença de competição referencial (esta última como constante nas imagens) na descrição de imagens, e nas expressões referenciais produzidas como parte das descrições. Comparo os dados obtidos para o Português Europeu com os dados para o Inglês obtidas na experiência de Coco e Keller (2015). A animacidade do referente teve o maior impacto nos resultados, levando a descrições iniciadas mais rapidamente, estruturalmente mais verbais, e, para o Português, com um maior número de expressões referenciais reduzidas. O Português e o Inglês diferiram também na estrutura das descrições, com o Português recorrendo a um maior nível de informação estrutural (com maior uso de Conjunções e Proposições) e o Inglês a mais informação referencial (com maior uso de Nomes e Pronomes). Não foram encontrados efeitos para o nível de visual clutter, talvez devido à natureza guiada da tarefa, e às variáveis testadas (que diferem de literatura prévia que estudou o nível de visual clutter na produção de expressões referenciais). De uma forma geral, os resultados reforçam o impacto da animacidade referencial na produção de expressões referenciais e das estruturas línguísticas envolventes, e mostram que este impacto é parcialmente influenciado pelas estruturas específicas da língua em uso.This thesis researches visually situated language production, with a focus on the production of referring expressions and a cross-linguistic approach that compares data for European Portuguese with data for English. Previous literature in visually situated language has studied the production of referring expressions with a basis on the idea of referential accessibility (e.g. Ariel, 2014; Gundel, Hedberg, & Zacharski, 1993): reduced or more informative referring expressions reflect referents that are more or less cognitively accessible. Based on this idea, there has been a search for the factors that influence referential accessibility. In this thesis, I focus on three factors that have been previously studied in visually situated language: referent animacy, the level of information of the visual context, and the presence of referential competition. Effects have been found in the production of referring expressions for all three of these factors (Arnold & Griffin, 2007; Clarke, Elsner, & Rohde, 2013; Fukumura & Gompel, 2011; Koolen, Krahmer, & Swerts, 2015), but they have been studied in isolation, and with different aspects of the referring expressions under analysis. Through a cued language production experiment, I test the impact of referent animacy, the level of information of the visual context (both as independent variables), and the presence of referring competition (as a a constant and integral property of the stimuli) on the description of images, and on the referring expressions produced as part of the descriptions. I compare the data for European Portuguese obtained in this experiment with the data for English obtained in the experiment of Coco and Keller (2015). Referent animacy had the largest impact in the results, with animate referents leading to more quickly initiated descriptions that are more verbal in their structure, and, for Portuguese, in a greater number of reduced referring expressions. Portuguese and English also differed in description structure, with Portuguese participants using a higher degree of structural information (through the use of Prepositions and Conjunctions), while the English participants make more use of referential information (with an increased proportion of Nouns and Pronouns). No effects of visual information were found, maybe due to the guided nature of the task, and the specific tested variables (that differ from previous literature which looked at the impact of visual context information on referring expression production). Overall, the results support the high impact of referent animacy on both referring expression production and on the surrounding linguistic structures that has been previously found in literature, while showing that this impact is partially influenced by specific properties of the used language.porLíngua portuguesa - Gramática comparada - Língua inglesaLíngua inglesa - Gramática comparada - Língua portuguesaPercepção visualLíngua portuguesa - Discurso (Linguística)Língua inglesa - Discurso (Linguística)Linguística cognitivaTeses de mestrado - 2018Shared mechanisms in situated language production : a cross-linguistic comparison between Portuguese and Englishmaster thesis202177980