LA TRADUZIONE LETTERARIA SERBO-ITALIANA CON CHATGPT. Limiti, potenzialità e prospettive

The present paper explores the potential and limitations of the generative model ChatGPT in the field of literary translation from Serbian into Italian. Building on recent studies that have highlighted the critical issues surrounding AI models, especially in the case of low-resource languages (Hendy et al. 2023), as well as the importance of the interpretive and creative dimension of translation – which goes beyond the mere application of rules (Marino 2022) – the paper analyzes a number of translation examples drawn from Momo Kapor’s novel Beleške jedne Ane (1972). At the lexical level, the analysis focuses on slang, idiomatic expressions, and realia, with particular attention to cases involving diachronic variation. At the phonetic level, it considers passages in which rhyme and sound patterning are woven into the narrative texture, thereby generating specific translation challenges. The aim is to identify the main issues that arise when generative models are employed in the literary domain and to outline possible strategies based on the use of different prompt types and on the evaluation of their impact on the stability and adequacy of the generated translations. The findings may prove useful for researchers, professional translators, and teachers interested in exploring the potential of generative translation and assessing its effectiveness compared with traditional translation approaches and established digital tools.