earticle

영어

Machine translation (MT) output falls short of human translation (HT) in terms of accuracy and readability, despite continued technological advances in its field. Against this backdrop, significant attention has recently been directed toward machine translation post-editing (MTPE), where a text pre-translated by an MT engine is post-edited by a human translator to meet client needs. In this study, a text translated from English to Korean via MT was post-edited, and the post-edited output was analyzed to identify any errors that were not properly corrected. The accuracy of the post-editing process was then quantitatively analyzed against HT of the same text. The HT group was found to spend approximately 1.5 times more time on average compared to the MTPE group. The post-editing accuracy, i.e., the ratio of corrected errors to the total number of errors, tended to increase as more time was spent on the task. The average post-editing accuracy of the MTPE group was 39.6%, significantly lower than the translation accuracy of the HT group at 73.6%. This suggests that even trained individuals find it challenging to thoroughly identify and correct accuracy errors in densely packed, machine-translated technical texts.