Post-editing Machine Translation in MateCat: a classroom experiment

Abstract

Advances in machine translation resulted in an increase of both volume and quality of machine-translated texts. However, machine translation still requires humans to post-edit the translation. This paper proposes a product-based approach of a post-editing (PE) experiment that was carried out with a total of 10 MA translation students. The goal of this study comprised both the analysis of the post-editing results performed by student translators involving a machine-translated text in MateCat and the subsequent error markup. By comparing the quality reports obtained at the end of the post-editing process, we analysed the linguistic quality results and observed a heterogeneous error distribution, considerable divergence in severity level ratings and a huge span of TTE (time to edit). This study aims at making a contribution to the integration of post-editing activities into the translation technology classroom for students without prior experience in PE.