Assessing Machine Translation of Emotional Depth, Metaphorical Complexity, and Cultural Nuances in Literary Texts — A Case Study of Dream of the Red Chamber

Abstract

Machine translation (MT) has significantly evolved, shifting from rule-based methods through statistical approaches to contemporary neural machine translation (NMT). Despite these advancements, translating literary texts—especially poetry and prose rich in emotional depth, metaphorical complexity, and cultural nuances—remains challenging. This qualitative comparative textual analysis investigates MT’s effectiveness in translating emotional expressions, metaphors, imagery, and culturally specific references from Chinese to English of selected excerpts of the classical Chinese novel Dream of the Red Chamber. Three popular AI translation systems—DeepL and Google Translate (specialized neural machine translation platforms) and DeepSeek (a generative large language model with translation capabilities)—were evaluated on their capability in translating texts of emotional depth, metaphorical complexity, and cultural nuances. The findings indicate substantial differences among the evaluated systems. DeepSeek consistently demonstrates superior performance across all dimensions, effectively capturing emotional subtlety, metaphorical depth, and cultural nuances. Google Translate provides translations of moderate quality, accurately conveying core meanings yet lacking nuanced literary and cultural resonance. DeepL, conversely, is faced with significant challenges, frequently resulting in awkward phrasing, distorted metaphors, and diminished emotional quality. These results align with the existing literature highlighting persistent limitations of MT in literary contexts. Future research should explore enhanced literary training datasets and emotion-aware MT techniques to bridge these identified gaps.