engleski

Training in post-editing for a changing profession

A recent language industry survey (ELIA et al. 2018) has shown that nearly 70% of EU translation companies and individual translators employ machine translation (MT) in their workflows. This trend has been made possible by recent improvements in the quality of MT output, mainly due to the development of a new, neural type of MT. The improved output is much more fluent, particularly when it comes to morphologically rich languages, but this very fluency can make some errors it contains all the more difficult to spot. Human correction of errors in MT output, known as post-editing (PE), is consequently an important new skill, increasingly seen as indispensable for translation professionals and arguably also becoming a necessary ingredient in translator education. The aim of this paper is to investigate the impact of a training course in post-editing for graduate students of translation on their PE performance. An earlier study (authors, forthcoming) has shown that without such training students tend to overlook a considerable number of errors in MT output, especially when working under constraints such as time pressure. A group of students starting a PE course will be asked to post-edit the Croatian translation of an English specialized source text produced by Google Translate. A similar experiment will be conducted with a comparable text at the end of the course in order to assess the expected improvement in their performance. Their results will also be compared to those of their peers who did not receive dedicated training in PE. The findings of this study are expected to inform curriculum design, while also providing more general insights into post-editing skill and its acquisition.

translator education ; post-editing ; machine translation ; translation technology

nije evidentirano