Relações coesivas em ambiente de produção textual multilíngue

Abstract

This thesis presents a study of COHESIVE RELATIONS established in a text written in Brazilian Portuguese and in four METATEXTS produced in English based on the Portuguese source text. The study draws on Systemic-Functional Theory (HALLIDAY; HASAN, 1976, 1985; HALLIDAY; MATTHIESSEN, 2014; FIGUEREDO, 2007, 2011) and Multilingual Studies (MATTHIESSEN et al., 2008). The set of texts examined is made up of a doctoral dissertation abstract in Brazilian Portuguese Nuclear Engineering area and its English version; two versions of the Brazilian Portuguese abstract translated into English under experimental conditions by a professional translator and a domain specialist respectively; and an abstract of an article written in English by the doctoral dissertations author based on the dissertation and published in an international scientific journal. The study aimed at investigating COHESIVE RELATIONS (HALLIDAY; HASAN, 1976, 1985) in order to map the selections made in each text within the English and Brazilian Portuguese SYSTEMS and to examine proximity relations among the texts, based on the types of resources selected. A further aim was to compare the COHESIVE RELATIONS results with those obtained by Silva (2016) for the textual SYSTEM OF MESSAGE. The methodology included (1) identifying and annotating cohesive categories using a COHESION annotation tool; (2) calculating annotation frequencies; (3) analyzing Rhetorical Relations (MATTHIESSEN, 2004) to complement conjunction analysis; (4) processing data for cluster analysis and dendrogram generation in R (R CORE TEAM, 2017); and (5) comparing the results for the SYSTEM OF COHESION with Silvas (2016). Cohesive items were classified drawing on Figueredo (2007, 2011) and Pagano et al. (2015) for Brazilian Portuguese, Halliday and Hasan (1976, 1985) and Halliday and Matthiessen (2014) for English. Based on quantitative data on the annotated categories, dendrograms were obtained, which show proximity relations among the texts. The article abstract published in an international journal formed a separate branch in the dendrogram with regard to identity and conjunctive chains, confirming Silvas (2016) findings. Considering identity chains, it was observed the clustering of the source text in Brazilian Portuguese with the doctoral dissertation abstract and the text translated by a domain specialist. With regard to similarity chains, dendrograms showed the source text in Brazilian Portuguese in a separate branch from the other texts, a result ascribable to the lower number of repetitions and higher number of instances of synonymy and collocation. Finally, regarding conjunctive chains, the source text in Brazilian Portuguese clustered with the doctoral dissertation abstract in English, the text translated by the professional translator and the text translated by the domain specialist.