MACHINE TRANSLATION POST-EDITING COMPETENCE IN THE AGE OF GENERATIVE AI: A CORPUS-BASED INVESTIGATION OF STUDENT TRANSLATORS' SKILLS

Abstract

The main aim of the article is to investigate how translation students interact with AI-generated translations and examines the relationship between AI literacy and post-editing performance. Using a corpus-based mixed-methods approach, the study compares human translation, neural machine translation post-editing, and generative AI post-editing. The findings reveal that generative AI improves fluency and lexical diversity but introduces AI-specific challenges, including hallucinations and contextual distortions. Statistical analysis demonstrates that AI literacy, critical evaluation skills, and prompt formulation abilities significantly influence post-editing quality. The study proposes an AI-enhanced MTPE competence framework integrating linguistic, technological, evaluative, and collaborative dimensions. The results contribute to current discussions on translator competence and provide practical recommendations for translator education in AI-mediated environments. The study employs a mixed-methods research design combining corpus linguistics, quantitative statistical analysis, and qualitative thematic analysis. Data were collected from 60 undergraduate and graduate translation students divided into three groups: Human Translation (HT), Neural Machine Translation Post-Editing (NMTPE), and Generative AI Post-Editing (GAI-PE). The learner corpus consisted of approximately 90,000 words. Translation quality, error frequency, lexical diversity, and AI literacy indicators were analyzed using corpus tools and statistical procedures. The study expands traditional translation competence models by incorporating AI literacy, prompt engineering, hallucination detection, and human–AI collaboration as essential dimensions of MTPE competence. Unlike previous studies focusing mainly on translation quality and productivity, this research examines how AI-related competencies influence student translators' performance in generative AI environments. The findings indicate that generative AI enhances translation fluency and lexical richness while simultaneously creating risks associated with factual inaccuracies and hallucinations. AI literacy emerged as a significant predictor of post-editing performance, highlighting the growing importance of critical evaluation skills in translator training. The study concludes that future translator education should integrate AI literacy, prompt engineering, and critical assessment strategies to prepare students for technology-rich professional environments.