An Empirical Study on English-to-Chinese Translation Quality: A Comparative Analysis of DeepSeek V3 and Doubao in Energy Technology Texts

Abstract

The rapid advancement of Large Language Models (LLMs) has increasingly highlighted the value of Artificial Intelligence (AI) translation in specialized domains. Sustainability reports in the energy sector, as typical informative professional texts, exhibit characteristics of terminology density, logical rigor, policy orientation, and interdisciplinary integration. Their translation quality directly impacts the effectiveness of technical exchanges, carbon reduction cooperation, and global energy governance among multinational energy enterprises, thereby necessitating stringent standards for translation professionalism, accuracy, and terminology consistency. However, existing research predominantly focuses on AI translation quality assessment in literary, financial, and other fields, while empirical studies on niche texts such as energy sustainability reports remain scarce, leaving industry demands for specialized translation solutions largely unmet. In response, this study selects core corpora from Enel’s 2023 Sustainability Report, employs DeepSeek V3 and Doubao as research subjects, constructs a comprehensive evaluation framework of “parallel text comparison and human review”, and systematically compares their English-to-Chinese translation performance across four dimensions,which are terminology processing accuracy, syntactic transformation rationality, logical coherence integrity, and professional context adaptability.