Role-based Prompting for Translation: A Study on Language Style and Translation Technique of Airline Slogan Translation Generated by ChatGPT

Abstract

A recent popular topic of interest in the study of ChatGPT as an AI translation technology is the quality of translations generated by ChatGPT and the factors influencing this AI to meet users’ expectations. However, translation studies studies focusing on role-based prompting for translation are still rarely under investigation, particularly those involving advertising materials such as airline slogans. This study explores a novel prompting strategy for ChatGPT by assigning role-based prompting utilizing persona pattern to translate English airline slogans into Indonesian. Airline slogans were purposely chosen due to their typical memorability, differentiation, and emotional connection to retain global customers in the aspects of safety, comfort, or unique services. It aims to identify the typology of language style and translation techniques emerged from the translations generated by ChatGPT. These two aspects are significant to study due to slogans’ typical syntax structures and linguistic choices that build them up and the need to improve understanding of AI’s translation behavior when translating slogans. Under the design of qualitative research, four stages of the prompt design stage; ChatGPT’s translation generation stage; data coding stage; and analysis on language style and translation technique stage were conducted by the researchers. First, ChatGPT 4.0 was commanded to translate 30 data of English international airline slogans collected from various online information sources, including the airlines’ official websites, into Indonesian using the “skilled translator” prompt (Prompt 1) and “translation professor” prompt (Prompt 2), which basically represents two expert professions from industry context and academic context. Their Indonesian translations generated from ChatGPT 4.0 were then coded and examined using translation comparative methods. To identify the types of language style and translation techniques appearing from the perspective of each role, Grey’s (2008, in Munder, 2024) theory of language style types, Shi’s (2014) linguistic components of slogans, and Li’s (2025) five techniques of translating slogans were applied. The findings reveal that when commanded with Prompt 1 and 2, ChatGPT generates outputs reflecting the viewpoint of each role. Prompt 1 produces shorter, catchier phrases and exhibits some colloquial lexical choices suiting advertising needs although they tend to use more literal translation techniques. Prompt 2 leans towards semantic fidelity using slightly longer words and more alliterative touch in a grandiose manner to likely achieve poetic appealing. Unexpectedly, instead of employing creative translation techniques to render the source texts, under both prompts, ChatGPT applies literal translation and free translation as its predominant techniques. This study contributes theoretically to providing more empirical studies on role-based prompting for translation and optimizing ChatGPT prompt training to improve ChatGPT’s creativity in translating advertising materials. Practically, it assists ChatGPT’s everyday users to efficiently generate real-time outputs nearly meeting their expectation. Nevertheless, reassuring the ChatGPT’s outputs in line with the airlines’ value and identity behind the slogans still needs to be done.