神经网络机器翻译的英译对比研究——以拉祜族创世史诗《牡帕密帕》为例

A Comparative Study on English Translation by Neural Machine Translation—Taking Mupamipa, the Creation Epic of the Lahu Ethnic Group, as an Example

摘要: 随着“一带一路”与中国文化“走出去”的推进，民族典籍英译地位凸显，翻译质量受学界关注。NMT技术的发展为其翻译质量的提升提供新可能。文章借BLEU和ROUGE指标，对比《牡帕密帕》谷歌英译文与人工译本，评估NMT的应用效能。结果显示，谷歌翻译在效率、通用句式处理具备优势，但在标题格式、上下文理解、术语一致性及文化内涵存在不足，BLEU与ROUGE评分低，与人工译本差距大。这表明当前NMT在民族典籍英译中适用性有限，需通过技术赋能与人工翻译辅助的双重保障，以满足准确性与文化传递的需求。

Abstract: With the advancement of the “Belt and Road” Initiative and the “Going Global” strategy of Chinese culture, the English translation of ethnic classics has become increasingly prominent, and its translation quality has attracted academic attention. The development of NMT technology provides new possibilities for enhancing the quality of such translations. By using BLEU and ROUGE as evaluation metrics, this paper compares the English version of Mupamipa translated by Google Translate with its human-translated version, so as to assess the application effectiveness of NMT. The results show that although Google Translate has certain advantages in translation efficiency and handling of common sentence structures, it has obvious shortcomings in aspects such as title formatting, context comprehension, terminology consistency, and cultural connotation. Consequently, its BLEU and ROUGE scores are generally low, indicating a significant gap between machine-translated texts and high-quality human-translated versions. This suggests that the current applicability of NMT in the English translation of ethnic classics is limited, and dual guarantees of technological empowerment and human translation assistance are still necessary to better meet the requirements of accuracy and cultural transmission in the English translation of ethnic classics.