Translating the same text twice

Abstract

This article presents the results of a study in which students in a graduate translation technologies course post-edited a text they had previously translated earlier in the semester without using machine translation (MT). The results show that post-editing allowed students with performance levels below, at, and just above an established median to improve the quality of their translation products, while students with performances well above the median actually experienced a decrease in quality. Nevertheless, the post-edited products and post-editing performances of the latter remained superior to those of the former. The study shows how different translators experienced gains or not in quality by accepting different aspects of MT output and how the accepted output relates to their human renditions. It also tracks whether their post-edits were necessary and correct and how they relate to their human renditions. Tracking such behaviors attempts to provide a more holistic view of how post-editing might be qualitatively advantageous or disadvantageous.