International Open Access Journal Platform

大语言模型辅助文化遗产文本英译策略研究——以《山海经》翻译中提示词与译后编辑路径为例

Strategies for Large Language Model-Assisted Translation of Cultural Heritage Texts —A Case Study of Prompt Engineering and Post-Editing in Translating Classic of Mountains and Seas

吴彦菲

Progress in Social Sciences / 2026,8(4): 314-321 / 2026-05-06 9 5

View Download PDF

• Information:

  华东理工大学，上海

• Keywords:
  Human-AI collaborative translation; Human-AI collaboration; Post-editing; Translation of cultural heritage
  人机协同; 提示词优化; 译后编辑; 文化遗产文本外译
• Abstract: Against the backdrop of rapid AI advancement and the growing demand for the cultural classics going abroad, this paper explores the applications and limitations of Large Language Models (LLMs) in translating the Classic of Mountains and Seas under a human-AI collaborative translation (HACT) framework. To this end, building on prior research frameworks, the study attempted to develop a three-tiered, progressive prompt system. By comparing outputs from prompts of different levels of detail, the study finds that refined prompt engineering significantly enhances translation quality. However, AI still lags behind expert human translators in conveying culture-specific items, achieving literary sentence structures, and capturing the underlying aesthetic essence of the text. Deep human post-editing is essential to refine terminology, rhythm, and discourse logic. Ultimately, this paper argues that maximizing the potential of LLMs in translating ancient classics requires a synergy between multi-layered pre-translation prompting and expertled post-editing. This collaborative approach provides a practical model for the cross-cultural transmission of intangible cultural heritage. 在人工智能技术迅速发展与文化典籍对外传播需求日益增强的背景下，本文探讨了在人机协同的翻译模式下，大型语言模型在《山海经》英译中的应用与局限。为此，研究基于前人研究框架尝试构建了一套针对文化遗产文本翻译的三层级递进式提示词。通过对比不同程度的指令对同一原文的翻译产出，研究发现精细化的提示词设计能显著提升机器译文的整体质量。但在文化专有项的准确传递、句式的文学性转换以及整体意蕴的再现方面，AI译文仍与高水平的人工译文存在差距。后续的人工深度编辑能够有效弥补这些不足，优化术语、节奏与篇章逻辑。本文认为要充分发挥大语言模型在典籍翻译中的潜力，关键在于将译前多层次、有针对性的指令工程与译后基于专业素养和审美判断的深度加工相结合，形成高效的人机协同路径，为文化遗产文本的跨文化传播提供实践示例。
• DOI: 10.35534/pss.0804056 (registering DOI)
• Cite: 吴彦菲．大语言模型辅助文化遗产文本英译策略研究——以《山海经》翻译中提示词与译后编辑路径为例［J］．社会科学进展，2026，8（4）：314-321.