1) Associate Professor, Department of Translation Technology and Practice at AKM-WEST, Northern (Arctic) Federal University, Russia, Arkhangelsk, m.berendyaev@narfu.ru 2) Assistant Professor, Department of For-eign Languages and Communication Technologies, College of Basic Pro-fessional Studies, MISIS University of Science and Technology, Russia, Moscow, m.gilin@misis.ru 3) Ph. D. of Philology, Docent, Head of Department of Translation Technology and Practice at AKM-WEST, Northern (Arctic) Federal University, Russia, Arkhangelsk, e.s.kokanova@narfu.ru e.s.kokanova@narfu.ru
The paper reviews approaches to evaluation of translation output in the context of using machine translation and automatic text generation systems. A new metric for assessing the quality of automatically generated text based on the predicted distance of its post-processing is proposed. The metric is built around the labor intensity of error correction and the risks of error impact on achieving the goals of a translation project. This metric aims to address the problems of evaluating the quality of auto-matic translation that have emerged with the advent of generative artificial intelli-gence replacing machine translation.
evaluation of translation output; translation quality assessment; machine translation; automatic text generation; metric; predicted post-process time; generative artificial intelligence.
Download textFor citing: Berendyaev M.V., Gilin M.I., Kokanova E.S. (2025) Generative AI and evaluation of translation output. Human being: Image and essence. Humanitarian aspects. Moscow. INION RAN.Vol. 2 (62). pp. 173-186. DOI: 10.31249/chel/2025.02.10