AI 기반 한국문학 번역 전략 탐색 ― Google AI Studio의 Temperature 조절과 프롬프트 설계를 중심으로

Abstract

This study employs Google AI Studio, a large language model platform that allows relatively intuitive adjustment of Temperature values, to examine how Temperature settings influence translation output and how the results change when such adjustment is combined with prompts that explicitly instruct particular translation strategies. The analysis demonstrates that lower Temperature values tend to produce translations that more closely adhere to the source text, whereas higher Temperature values generally yield more contextually appropriate lexical choices and more natural syntactic structures. With respect to metaphorical and simile expressions, the study finds that explicit guidance on how to interpret and render these expressions, combined with a higher Temperature setting, enables the model to more faithfully convey subtle semantic nuances and to produce more natural choices in both vocabulary and syntax. Drawing on these findings, this research explores how AI can be methodologically integrated into literary translation and argues that AI is best positioned as a supplementary tool, not a replacement for the human translator.