Abstract:
Medical translation in high-risk domains such as healthcare requires workflows that combine efficiency with rigorous quality assurance. This article proposes a hybrid model integrating Automatic Post-Editing (APE) and Translation Quality Assessment (TQA) to improve English-Romanian medical translations. Building on ISO 18587 and error taxonomies such as Multidimensional Quality Metrics (MQM) and Dynamic Quality Framework (DQF), the framework introduces APE as an intermediate error-reduction layer and hybrid TQA combining human annotation with automated quality estimation. The study proposes an eight-stage workflow aligned with international standards and tailored to risk-sensitive genres like patient information leaflets and discharge summaries. In addition, a theoretical pilot protocol is presented to validate the workflow and evaluate accuracy, fluency, and efficiency across three translation pipelines: human translation, Neural Machine Translation (NMT) with post-editing, and NMT with APE plus post-editing. By bridging academic research and professional practice, this contribution advances a standards-driven, implementable approach for technology-mediated medical translation, addressing Romanian-specific gaps in resources, training, and workflow design.