5. Modelling proximity in a corpus of literary retranslations: A methodological proposal for clustering texts based on systemic-functional annotation of lexicogrammatical features

Abstract

This chapter seeks to contribute to a model for quantitative exploration of translated texts by adopting clustering techniques to search for patterns of comparability in a corpus of retranslations. Drawing on systemic-functional theory (Halliday & Matthiessen, 2004) as a framework for text analysis, it reports on an exploratory study aimed at investigating source – target text relations as computed through statistical methods for a manually annotated representative text sample. Annotation built on the analytical framework used for comparing source and target texts, which is based on categories of grammatical functions common to both source and target language systems. The corpus was compiled from ten translations of a source text -- a short story written in English by Katherine Mansfield -- into Spanish and Portuguese by different translators over a period of six decades. The texts were explored in terms of the ‘retranslation hypothesis’ (Berman 1990), whereby retranslations tend to be more source-oriented than first translations, orientation being established in our study on the basis of the distance between source and target text as computed through cluster analysis. The results obtained point to similarities between texts computed on the basis of categories ascribed to the lexicogrammatical choices made by each author within the grammatical systems analyzed. They also corroborate the findings of other researchers who have used other approaches and methodologies to probe the ‘retranslation hypothesis’, in that they confirmed the relative distance of a first translation from the source text, while they also showed varied degrees of proximity of retranslations to the source text, the former being in some cases further away from the latter than first translations.