Prompting ChatGPT for Translation: Lexical and Syntactical Customization in Translation of “The Tree Rhyme”

With the advancement of ChatGPT, scholars are delving deeper into the tool's ability to generate various texts, including literary texts. It becomes potential due to the lack investigations on prompting ChatGPT for translating children’s nursery rhymes, particularly in Indonesia, making to the novelty of this study. This article demonstrates how ChatGPT reacts to different prompts when translating a nursery rhyme “The Tree Rhyme”, by concentrating on the lexical and syntactical customization. This preliminary investigation was conducted in three phases: 1) prompt design, which includes one basic prompt functioning as a baseline for comparison and three other prompts functioning as clues to explain the case of linguistic customization; 2) text generation, which takes a level-2 nursery rhyme book entitled “The Tree Rhyme” from a digital repository, StoryWeaver; and 3) linguistic customization evaluation, which was done by comparing ChatGPT’s translations in response to the four provided prompts and observing the lexical and syntactical customization. The findings reveal that ChatGPT's lexical customization includes changing the last word of stanzas, reducing the number of syllables, subtracting words, and displaying child-friendly expressions. These modifications cannot result in translations that maintain the original text's rhyme scheme. Regarding the syntactical customization, ChatGPT generates concise structures containing aspects of children's literature by reordering, unit shifting, sentence simplification, direct construction, and phrase simplification, but still unable to replicate the ST’s rhyme scheme. In conclusion, ChatGPT changes diction and simplifies sentences when reacting to cues to translate the nursery rhyme, but it cannot preserve the ST’s rhyme scheme.