MACHINE TRANSLATION POST-EDITING TECHNOLOGIES

Abstract

The article examines machine translation post-editing (MTPE) technologies in technical documentation translation, with particular attention to productivity, translation quality, and translators’ professional experience. The relevance of the study lies in the growing use of MTPE in technical communication, where dense terminology, complex syntax, and functional precision make translation errors especially consequential. The research is based on a mixed-methods design that combines quantitative analysis of post-editing performance with qualitative investigation of translators’ decision-making and perceptions. Eleven professional translators working with the English – Ukrainian and English – French language pairs post-edited authentic segments from software user guides, API documentation, and troubleshooting articles. The data included time spent per segment, editing patterns, and final translation quality assessed by independent evaluators. Semi-structured interviews further revealed how translators identify errors, handle technical terminology, and balance productivity and quality. The findings demonstrate that post-editing efficiency varies significantly by content type: API documentation proved the fastest to process, while troubleshooting materials required the greatest effort and, in some cases, were slower to post-edit than to translate from scratch. Although most post-edited segments reached acceptable quality, persistent problems included terminological inconsistency, fluency issues, and subtle semantic distortions. The study also identified MT-induced errors that remained unnoticed because machine-generated solutions appeared plausible. A particularly important result is that technical domain knowledge strongly influences both productivity and quality, as translators with relevant subject expertise performed more effectively and reported greater confidence in MTPE workflows. The article concludes that post-editing in technical translation should be understood not as a purely mechanical correction of machine output, but as a cognitively demanding and professionally complex activity that requires linguistic competence, technological awareness, and domain-specific knowledge.