A Study on Translation Quality Assessment of English Versions of the Shang Han Lun Based on Corpus Linguistics and Large Language Models

Abstract

Driven by the dual national strategies of TCM internationalization and cultural digitization, high-quality English translation of TCM classics has become a core component of TCM cross-cultural communication. However, the quality of existing English translations of TCM classics varies greatly, and there is a lack of quantifiable and reusable evaluation standards. Traditional evaluation models often rely excessively on experts' subjective judgments, which hinders the scientific selection of translations and the improvement of communication effectiveness. Taking the core text of Shang Han Lun as the research object, this study selects five translations (Luo Xiwen, Wiseman, Cheng Zhaozhi and Chen Jiaxu, Greta Young Jie De, Huang Hai) to construct a Chinese-English parallel corpus, and uses the GPT-4o large language model to generate reference translations. By integrating automated evaluation metrics with in-depth case analysis, this study constructs a composite evaluation system covering three dimensions: accuracy, fluency and domain adaptability. Empirical results show that there are significant differences among the translations in terminology consistency and the handling of culture-loaded words. Among them, Greta Young's translation has the highest comprehensive score, showing the best fit with the reference translation in terminology standardization and semantic transmission, and demonstrating strong potential for standardized dissemination. At the methodological level, this study couples the quantitative advantages of corpus linguistics with the semantic understanding capabilities of large language models, providing a reusable intelligent research paradigm for the translation quality assessment of TCM classics.