Large language models in translation quality assessment: The feasibility of human-AI collaboration

.This research explores the potential application of Large Language Models (LLMs) in translation quality assessment within the Chinese Academic Translation Project (CATP), from a human-AI collaboration perspective. The study integrates the LISA QA Model and the Chinese standard GB/T 19682-2005 to develop a multidimensional translation quality assessment system, including typologies and weights of errors specific to Chinese academic works. Using this system, three LLMs (GPT-4, Claude-3.7, and Deepseek-R1) were employed to evaluate the Portuguese version of the work Introduction to Qing Dynasty Academic Thought, analyzing their performance and comparing it with the results of an assessment conducted by human experts, with the aim of exploring the feasibility of a collaborative model between humans and AI. Based on the experimental results, the research proposes a hierarchical assessment process of “AI screening-refined human judgment” and an inter-linguistic assessment mechanism of “Chinese prompt-multilingual verification”, constructing a translation quality assessment framework based on human-AI collaboration for the CATP. This study infuses elements of technological innovation into traditional translation quality assessment, providing a new technical support pathway for the strategy of “internationalization” of Chinese academic knowledge.