Traduction humaine et traduction automatique de Grândola, Vila Morena en français

Abstract

José Afonso's poetic song, Grândola, Vila Morena, became 'the anthem of the Portuguese protest movement' (Ciccia 2013) and 'the anthem of the Carnation Revolution' (Gomez 2017). This revolutionary song has inspired numerous spontaneous translations into French. A sample of translations available online enabled us to compile a corpus of five human translations that highlight diverse interpretations of the original text and choices of convergent or divergent translation strategies. To these five versions, we added five versions produced by neural machine translation engines: one from Google Translate (January 2026) and two from DeepL, 'the world's best translator' according to its creators, and from eTranslation, the engine of the European Union's Directorate-General for Translation, generated in November 2024 and January 2026. This study aims to compare these ten versions in terms of denotative and formal-aesthetic equivalence (Koller, 1992) to identify the translation strategies employed by the human translator or the machine to address the challenges of this poetic text. The results show that denotative equivalence is prioritised in human and machine-generated literal, verse-by-verse translations, with one exception. They also reveal that translation solutions are numerous and that semantic equivalence errors do not result solely from the machine's lack of common sense.