English Translation of Ancient Chinese Scientific Classics in the AIPE Era: Efficacy, Boundaries, and Pathways —An Empirical Study Based on the Translation Review of Mengxi Bitan

Abstract

Current research on Artificial Intelligence Post-Editing (AIPE) remains largely theoretical, lacking sufficient empirical, case-specific, and quantitative analysis. This study investigates the efficacy, boundaries, and developmental pathways of AIPE in translating ancient Chinese scientific classics, using the English translation of Mengxi Bitan (Brush Talks from Dream Brook) by Shen Kuo (1031-1095) as a case study and integrating quantitative and qualitative methods. The findings reveal that, despite multiple rounds of prompt engineering and Few-Shot Learning (FSL) training, AI continues to exhibit significant limitations: it is prone to factual and contextual hallucinations, lacks empathetic perception, and struggles to capture the emotional and cultural nuances of the source text or adapt to target readers’ cognitive gaps. The study concludes that AI currently serves only as an auxiliary tool and is not yet capable of functioning independently as a translation reviewer in the short term. This research aims to fill the empirical gap in AIPE-assisted translation of scientific classics, underscores the necessity of building a human-AI collaborative model, and proposes strategies for enhancing the competence of human reviewers and optimizing AI performance to achieve synchronous improvements in both translation efficiency and quality.