Exploring the Potential of ChatGPT for Translating Profanities in Film Subtitles: A Case Study of Parasite and Joker

This study explores the potential of ChatGPT for translating profanities in film subtitles, with a focus on the movies Parasite (Korean) and Joker (American). Profanity in subtitles presents unique challenges, as it not only conveys emotional intensity but also reflects cultural and social contexts that require careful preservation in translation. The research examines the effectiveness of ChatGPT by comparing translations generated under two types of prompts: (1) basic prompts requesting simple translation and (2) detailed prompts designed to capture cultural and emotional nuances. These outputs are compared alongside human translations using a mixed-method approach, combining quantitative metrics such as translation strategy frequency and qualitative evaluation of cultural and emotional fidelity. Findings reveal that ChatGPT with detailed prompts (GPT2) performs significantly better in maintaining the cultural and emotional context of profanities, often matching human translations in emotional depth. However, challenges remain, particularly in interpreting idiomatic expressions and capturing subtle cultural differences. GPT1, with its generic prompts, showed a tendency to neutralize or omit profanities, weakening the original emotional impact. The research highlights the importance of prompt specificity in improving AI translation outcomes and identifies areas where human oversight remains critical. By combining AI efficiency with human contextual expertise, this study envisions a collaborative translation model for future applications. The findings also underline the need for continued advancements in AI translation models to handle culturally specific and emotionally charged language more effectively. This research contributes to the understanding of AI’s role in translating emotionally and culturally sensitive language and suggests directions for integrating AI in professional translation workflows.