How can AI-assisted literary translation from Spanish into Chinese be improved?

Abstract

In the field of artificial intelligence (AI)-assisted translation, the growing use of large language models (LLMs) opens up new possibilities for machine translation. Nevertheless, their application to literary texts continues to pose challenges related to the preservation of style and cultural nuances. In this context, the present study examines how to improve the quality of literary translation from Spanish into Chinese through the use of LLMs and post-editing. To this end, a sequential explanatory mixed-methods approach was adopted, with an exploratory-descriptive scope and a comparative character. Four prompts of increasing complexity were designed, and the translations generated by AI were compared with a human reference translation. The results show that the use of LLMs, combined with human post-editing, can improve the quality of literary translation from Spanish into Chinese. On the one hand, this improvement depends on the formulation of detailed prompts that include contextual information, such as the translator’s role, the literary style of the text, and reference examples. On the other hand, it also depends on post-editing aimed at making the language sound more natural and adapting expressions to the target culture. Human intervention is recommended at three levels: adjustment of register and literary rhythm, cultural adaptation through the use of idioms and metaphors characteristic of Chinese, and correction of the excessive use of punctuation marks.