人机协同下的文化负载词翻译：AI局限与译后编辑路径探索

Human-AI Collaborative Translation of Culturally Loaded Words: Challenges for AI and Post-Editing Approaches

摘要: 在人工智能快速发展的背景下，机器翻译在跨文化传播中的应用愈发广泛。然而，文化负载词作为语言与文化的交汇点，仍是AI翻译系统的核心难题。本文以ChatGPT为研究对象，聚焦其在文化负载词翻译中的表现及译后编辑的调控作用。研究发现，ChatGPT虽能在语言层面实现较高的可译性，但在语义理解、文化意象解码和语用适配方面仍存在明显的局限性。这一不足源于模型对统计语料的依赖及对深层文化知识的缺乏。译后编辑不仅修正了AI译文中的偏误，更通过文化判断与语篇重构实现意义再造，使译文在准确性、文化契合度与传播效果之间形成动态平衡。研究表明，人机协同是实现文化再生产与语际互释的重要途径，为智能翻译的优化提供了可行路径。未来研究可从技术与认知两方面深化探讨：一是提升AI模型的文化语义识别与隐喻理解能力；二是揭示译后编辑中的认知机制与决策路径，从而推动智能翻译的文化深度发展。

Abstract: Against the backdrop of rapid advancements in artificial intelligence, machine translation has become increasingly prevalent in cross-cultural communication. However, culturally loaded words, as the intersection of language and culture, remain a core challenge for AI translation systems. This study focuses on ChatGPT, examining its performance in translating culturally loaded words and the role of post-editing in optimizing its outputs. Findings indicate that while ChatGPT achieves high translatability at the linguistic level, it exhibits notable limitations in semantic understanding, cultural imagery decoding, and pragmatic adaptation. These shortcomings stem from the model’s reliance on statistical corpora and its lack of deep cultural knowledge. Post-editing not only corrects errors in AI-generated translations, but also reconstructs meaning through cultural judgment and discourse adjustment, achieving a dynamic balance between accuracy, cultural appropriateness, and communicative effect. The study demonstrates that human-AI collaboration is a crucial mechanism for cultural reproduction and interlingual interpretation, providing a practical pathway for optimizing intelligent translation. Future research may deepen exploration from both technological and cognitive perspectives: first, enhancing AI models’ ability to recognize cultural semantics and interpret metaphors; second, investigating the cognitive processes and decision-making strategies involved in post-editing, thereby promoting the cultural depth development of intelligent translation.