A comparative study of human and machine translation of scientific texts from the perspective of the three-dimensional transformation theory of Eco-translatology——A case study of Introduction to the Defect Chemistry of Metal Oxides
生态翻译学三维转换视角下科技文本人机译文对比研究——以《金属氧化物中的缺陷化学》为例

摘要

科技文本专业术语密集、逻辑严密，具有较强的知识传播功能，对译文的准确性和专业性提出了较高要求。然而，目前以各类人工智能大模型为基础的机器翻译在科技翻译领域中的应用效果却待于改进。本文以材料化学著作《金属氧化物中的缺陷化学》（Introduction to the Defect Chemistry of Metal Oxides）引言译文为研究对象，以ChatGPT机器译文和赵鸣教授人工译文为比较语料，在生态翻译学“三维转换”理论框架下，从语言维、文化维和交际维三个层面对两种译文进行比较分析。研究发现，机器翻译在术语对应、信息保留和源语结构再现方面具有较强优势，能够较好地保持原文语言形式；人工翻译则更注重专业知识背景、学术表达规范及读者适应度，在概念解释、语篇重构和交际功能方面表现更为突出。两种翻译模式体现出不同适应与选择机制，具有较强的互补性。人机协同翻译将成为未来科技翻译发展的重要方向。

Abstract

Scientific texts are characterized by a high density of technical terms, rigorous logic and strong knowledge-transmission functions, which impose high requirements on the accuracy and professionalism of translation. However, the application of large-language-model-based machine translation in science and technology still needs to be further improved. This study takes the introduction section of Introduction to the Defect Chemistry of Metal Oxides, a classic work in the field of defect chemistry and dielectric materials, as the research material. The study compares the machine translation generated by ChatGPT with the human translation produced by Professor Zhao Ming. Based on the three-dimensional transformation theory of Eco-translatology, the two translations are examined from the linguistic, cultural, and communicative dimensions. The comparative analysis reveals that the ChatGPT translation performs well in rendering technical terms, preserving source-text information, and retaining source-language structures. As a result, the linguistic features of the original text are largely maintained. The human translation pays greater attention to disciplinary knowledge, academic conventions, and readers’ comprehension. It shows advantages in concept explanation, discourse reorganization, and the realization of communicative purposes. The comparison indicates that machine translation and human translation adopt different strategies of adaptation and selection. At the same time, they complement each other in many respects. Machine translation can improve efficiency and provide an initial translation draft, while human translators can further refine the translation through professional knowledge and contextual understanding. The study suggests that human–machine collaboration provides a feasible approach to improving both the efficiency and quality of scientific translation.