A Study on the C-E Classic Sayings Translation Quality Assessment of LLM DeepSeek-R1 in Understanding Xi Jinping’s Educational Philosophy

Abstract

This study sampled 30 classic sayings contained in the Understanding Xi Jinping’s Educational Philosophy as the research objects. It takes the official translation version Understanding Xi Jinping’s Educational Philosophy which was commissioned to Beijing Foreign Studies University by the National Textbook Committee Office as the reference translation, and the translations generated by the domestically advanced Large Language Model (the following is abbreviated as LLM) DeepSeek-R1 as the objects for translation quality evaluation. The study adopts quantitative evaluation and qualitative analysis as research methods. Through the collaborative analysis of two indicators, namely COMET (Cross-lingual Optimized Metric for Evaluation of Translation) and TER (Translation Edit Rate), it quantitatively evaluates the performance of the evaluation objects in semantic fidelity, and the surface edit distance difference. At the same time, it qualitatively analyzes the cultural-loaded equivalence of the evaluation objects. The study finds that the LLM DeepSeek-R1 can translate well with accurate semantics and suitable contexts in the translation of classic sayings. However, when dealing with expressions with profound cultural connotations, there are still problems such as semantic deviations, insufficient transmission of cultural images, and a measurable surface edit distance from the reference translation. This study provides empirical evidence for enhancing the textual translation quality of the LLM DeepSeek-R1 in the field of classic sayings translation; it also serves as an important reference for future enhancements of the model in this field and the expansion of its application scenarios.