Post-editing strategy optimization and performance evaluation based on DQF-MQM error analysis

Medical machine translation (MT) post-editing faces significant challenges regarding insufficient targeting and poor adaptability to long texts. To address this, this study proposes a hierarchical post-editing strategy integrating the Dynamic Quality Framework (DQF) and Multidimensional Quality Metrics (MQM). Unlike traditional passive correction methods, this study introduces a proactive closed-loop model featuring ‘identification, risk assessment, and control’. We utilize medical clinical reports to construct an error priority model and a propagation risk mechanism based on a BERT-Knowledge Graph architecture. Experimental results demonstrate that this strategy improves editing efficiency by 41.26% ( p < 0.05), achieves an average MT quality score of 89.17, and attains F1 scores of 93.85% and 89.26% for terminology and semantic error recognition, respectively. Notably, in 5,000-word texts, editing time was reduced by over 45%. The strategy shows high robustness, with manual intervention increasing by only 0.8–1.2% for every 0.5‰ rise in term ambiguity density. This study provides a reproducible, risk-controllable paradigm for high-stakes medical translation, balancing quality and efficiency.