基于《活着》苦难叙事的白睿文译本与豆包AI译本的对读批评

A Critical Analysis of the Narrative of Suffering in the Novel To Live Based on Michael Berry’s Translation and Doubao AI-Generated Translation

摘要: 本研究通过建构“语言层–超语言层–差异根源”三阶分析框架，以余华《活着》核心苦难段落(pp. 98~107)为样本，对白睿文人类译本与豆包AI译本进行文本细读、对读，对比人工智能与人类译者在余华《活着》原作风格及苦难美学转译中的本质差异。研究发现：语言维度揭示人类译本通过语法结构重构、术语中性化等策略实现零度写作等效，而AI因情感计算过载与直陈式叙述消解了原文写作风格；超语言维度表明人类译者以文化符号重构与隐喻具体化传递苦难美学的意蕴，AI则陷入字面解码陷阱导致文化内涵丢失；追溯差异根源发现人类译员有意识地保留叙事寓言，AI则因伦理判断失误导致意义降维。最终揭示豆包AI文学翻译两大瓶颈——情感识别判断困难、文化转喻能力不足，本研究据此提出“AI初译→人类校改→综合校准”的人机协作模型，为文学领域的机器翻译提供借鉴。

Abstract: This study constructs a tripartite analytical framework (linguistic/extra-linguistic/root cause) to conduct a close comparative analysis of Michael Berry’s human translation and Doubao AI’s translation of the core suffering passages (pp. 98~107) in Yu Hua’s To Live, revealing fundamental differences in conveying the original style and aesthetics of suffering. Linguistically, the human translation achieves zero-degree writing equivalence through syntactic reconstruction and terminological neutralization, whereas the AI’s emotional computation overload and literal narration break up the original style. Supralinguistically, the human translator transmits the aesthetic essence of suffering via reconstructed cultural symbols and concretized metaphors, while the AI’s literal decoding fails to capture cultural connotations. Tracing root causes, human translators consciously preserve narrative allegory, but the AI’s ethical misjudgment reduces meaning depth. Finally, the study identifies Doubao AI’s critical bottlenecks—deficiencies in emotional judgment and cultural-metonymic competence. Thus we propose an “AI Draft → Human Revision → Collaborative Calibration” collaboration model, offering a practical paradigm for enhancing machine translation’s literary quality.