Over the past few years, artificial intelligence and machine translation have revolutionized the field of professional translation—nonetheless, many still consider literary translation an insurmountable obstacle for non-human translators. This is why I have decided to delve into this relatively under-researched field of literary translation into Slovene using artificial intelligence and machine translation. For the purpose of this master’s thesis, I have analysed five randomly selected passages from children’s novel Matilda by Roald Dahl. The central section of the study explores translational abilities of ChatGPT and DeepL by comparing their best translations with one by Gemini and with two human translations: one by Bogdan Gradišnik (1990) and the other by anonymous literary translator (2025), while the second part focuses on the progress of ChatGPT and DeepL over the course of one year. Each analysis is divided into three parts: the first part identifies serious orthographic, grammatical and syntax, stylistic and semantic, as well as lexical errors, in the second part, we evaluate the adequacy of translations, while the third part assesses the style in translations, with an emphasis on the naturalness and the stylistic suitability of the analysed translations. The progress analysis has shown no significant improvement in DeepL translations, whereas ChatGPT has exhibited noticeable advancement, mainly at the grammatical and stylistic levels. The central comparative analysis has confirmed that both human translations consistently outperform all non-human translations in all examined aspects—especially in grammar and style. Moreover, none of the human translations contain major lexical errors featured in all non-human translations.
|
