丁玲 문학과 신경망(NMT) 번역 ― 여성적 서사의 기계적 수용 가능성 ―

초록

This study explores how neural machine translation (NMT) systems handle the politically and emotionally charged language of feminist literature, focusing on Ding Ling's early works Amao Girl, and A Small Room in Qingyun Lane. Through a comparative analysis of translations produced by ChatGPT, Claude, and DeepL, the paper examines how key elements such as female desire, bodily expression, silence, and resistance are rendered by AI. Using both quantitative metrics (COMET scores) and qualitative criteria (emotional fidelity, metaphor translation, cultural sensitivity, and subjectivity), the study highlights critical discrepancies among the models. While ChatGPT maintains contextual coherence and stylistic fluency, it often neutralizes expressions of desire or resistance. Claude shows ethical caution, softening subversive narratives, while DeepL emphasizes sensory vividness at the cost of misinterpreting cultural nuance. The findings reveal that AI translation systems—despite their linguistic sophistication—struggle to reproduce the emotional depth, symbolic layers, and political resonance of feminist texts. The study ultimately argues for the continued necessity of human translators, especially when engaging with emotionally charged and ideologically complex literature. It further suggests a hybrid model where AI assists in structural tasks, while human sensitivity ensures interpretive accuracy and cultural fidelity.

키워드