<![CDATA[Preserving humor effects that arise from flouting conversational principles in culturespecific texts presents specific difficulties when translating regional languages deeply rooted in linguistic and cultural variation. This type of humor depends on complex linguistic, cultural, and situational contexts, resulting in difficulty in replicating the same effect in the target language. This study investigates humor translation strategies and translation quality of two advanced AI language models, i.e., ChatGPT 4.0 and Gemini 2.5, in rendering culture-specific humor that deliberately violates Grice’s maxims of cooperation and Leech’s politeness principles. The study focuses on translations from Sundanese into English using 20 humorous texts selected from Cangehgar, a well-known Sundanese humor anthology characterized by deliberate violations of conversational and politeness principles. Chiaro’s four-option humor translation framework is employed to identify the strategies used, while Larson’s scales of clarity, accuracy, and naturalness are applied to evaluate translation quality. A descriptive qualitative approach is adopted, in which the selected humorous texts were purposively sampled based on their clear instances of flouting conversational and politeness principles. The findings indicate that while both models employ distinct translation strategies, their effectiveness in preserving the humor’s original intent varies. ChatGPT establishes stronger capability in maintaining the structure and essence of humor in the punchlines, whereas Gemini’s adaptations tend to dilute the comedic impact. This study contributes to the growing academic discourse on AI’s role in humor translation and its broader implications for cross-cultural communication.]]>
Copyrights © 2025
