Bilingual Evaluation of Tone and Emotional Nuance in Korean-English Translation: Human vs. AI
Yeeun Choi
21/07/2026
The act of translation requires navigating tensions between fidelity to source texts and accessibility for target audiences. This challenge is especially present in Korean-English translation, where there is substantial linguistic and cultural distance. In an era where AI translation systems grow ever more sophisticated, the question arises regarding how bilingual readers, who are well-positioned to assess both source and target texts, perceive and evaluate translation quality across human and AI-generated outputs. This paper investigates how Korean-English bilingual speakers evaluate tone, emotional nuance, and cultural meaning in translation, and what criteria they prioritize when assessing translation quality. Twenty bilinguals evaluated excerpts from the Korean folktale The Tiger and the Dried Persimmon, comparing a human translation with a ChatGPT-4o generated version through Likert-style ratings and open-ended responses. Findings suggest that participants tended to prioritize emotional authenticity and tonal fidelity over literal accuracy, and considered cultural appropriateness largely context-dependent. While AI translation was preferred for conveying character voice and humor, neither translation adequately rendered culturally-specific expressions such as aigoo or yayya. Notably, participants could not reliably distinguish AI from human translation, with only 60% recognizing AI involvement, suggesting that translation quality depends on how each approach handles specific challenges rather than whether it was produced by human or AI. This study further explores the sophisticated metalinguistic awareness bilingual readers bring to translation evaluation by demonstrating that their criteria extends beyond emphasis on semantic fidelity.