A comparative analysis of poetry translations from the perspective of translation realm: a case study of "Returning to Lingao at Night" (Lin Jiang Xian）
翻译境界论视角下诗歌译本对比探析——以《临江仙·夜归临皋》为例

摘要

人工智能大模型的出现给翻译领域带来重大变革。文章从翻译境界理论出发,用境界作为古诗英译的评价标准，以苏轼《临江仙·夜归临皋》为源本，对文心一言4.0、5.0的译本与许渊冲的译本对比评价，旨在评估文心一言在诗歌翻译特定领域的表现和能力。结果显示：横向比较，文心一言5.0和4.0在译意上有很大改进，但译味和译境还是不及许渊冲的译本。这说明诗歌翻译的深度与复杂性使其难以完全替代人工。未来，随着深度学习与多模态翻译技术的迭代，人机协作翻译前景更广阔，我们要正视人工智能带来的机遇和挑战。

Abstract

The emergence of large artificial intelligence (AI) models has brought profound transformations in the field of translation. Grounded in the theoretical framework of “Jingjie” (“translation realm”), this article adopts “Jingjie” as a criterion for evaluating English translations of classical Chinese poetry. Taking Su Shi’s ci poem “Lin Jiang Xian: Returning to Lingao at Night” as the source text, the study comparatively assesses translation outputs generated by Wenxin Yiyan 4.0 and 5.0 against the renowned translation by Xu Yuanchong. The objective is to systematically evaluate Wenxin Yiyan’s performances and capabilities specifically in poetic translation.
Results indicate that, in horizontal comparison, Wenxin Yiyan 5.0 demonstrates notable improvements over version 4.0 in terms of semantic fidelity (“translation meaning”), yet both versions are not as good as Xu Yuanchong’s rendition with respect to aesthetic resonance (“translation flavor”) and evocation of poetic aesthetic (“translation aesthetic”).This suggests that the profound artistic and linguistic complexity inherent in poetic translation render it difficult to full replaced by AI. Looking ahead, as deep learning techniques and multi-modal translation technologies continue to evolve, human–AI collaborative translation holds increasingly promising prospects. Scholars and practitioners need to adopt a balanced, critical perspective—acknowledging both the opportunities and challenges posed by AI in translation.