An Analysis of Indonesian Idiom Translation in Sapardi Djoko Damono's Novel Hujan Bulan Juni by ChatGPT and Gemini

Abstract

This research notes that despite the growing use of Large Language Models in translation, empirical research comparing how different AI systems apply specific translation strategies to handle culture-bound literary idioms remains limited. To address this gap, this research examined how ChatGPT and Gemini translate Indonesian idioms into English and analyzed how their strategies contributed to maintaining semantic equivalence. A descriptive qualitative approach was adopted, analyzing twenty-three idiomatic expressions extracted from Sapardi Djoko Damono’s novel Hujan Bulan Juni as the primary source. The data were evaluated using Baker's (2018) idiom translation framework, which was selected over alternative translation models due to its exceptionally comprehensive and systematic taxonomy in mapping both structural and functional equivalence across cross-cultural linguistic transfers. Emphasizing a comparative analysis of translation strategies rather than mere translation quality, the findings revealed that both ChatGPT and Gemini predominantly used translation by paraphrase, followed by maintaining similar meaning with different forms, as well as preserving both meaning and form. Translation by omission, however, was not detected at any point during the data analysis. The results also showed that ChatGPT tended to prioritize semantic clarity through paraphrase, whereas Gemini more frequently used equivalent English idioms. Both AI systems generally achieved semantic equivalence, although some cultural and figurative nuances were not fully preserved. Ultimately, this research contributes significantly to the field of AI-assisted literary translation by providing empirical insights into the methodological behaviors of generative AI models in handling figurative language.