Analysis of user post-edited texts and a proposal for assistance through iSTS

Abstract

In the present investigation, the way lay users employ machine translation (MT) systems has been studied. The benefits of applying these systems for the dissemination have been examined to limit the scope, and the Polish-Spanish language pair has been selected. Prior to analyzing the similarities and differences in the compositions that these users have created, with and without the assistance of MT, various state-of-the-art systems, along with their strengths and weaknesses, have been described. Moreover, some of the available MT evaluation methods that address those weak points have been presented. However, it is considered that there is still a gap when it comes to assisting lay users in post-editing. It is believed that interpretable semantic textual similarity (iSTS) could fill this void. Nevertheless, some refinements of its annotation guidelines might be necessary.