Overview of the Major Findings of the Szeged Post-Editing Research Project

Absztrakt

This paper reports on a multi-year research project carried out by the Translation and MT Post-editing Competence Research Centre of the University of Szeged, Hungary. The Centre conducted a complex project to investigate the role of different competence elements in traditional from-scratch human translation (FSHT) and in the post-editing of neural machine translation (MTPE). The project began in 2020 and relied primarily on the PACTE model of translation competence. Between 2020 and 2024, data were collected from 106 students, who translated or post-edited a research paper abstract or a legal text from English into Hungarian, completed language tests and background knowledge tests, filled in a form on their beliefs about translation and a follow-up questionnaire. Drawing on models of translation and postediting competence, we analysed how competence elements, source-text characteristics and task type (FSHT vs. MTPE) interact. In this paper, we attempt to synthesise the main findings of the project, with a particular focus on comparing FSHT and MTPE with respect to the effect of source text (ST) genre and domain, and on feedback from students working with MTPE. Overall, our findings suggest that post-editing offers clear advantages only when the genre, the content and the register of the ST are familiar to the translators / post-editors. An implication of these findings is that the development of language competence, particularly reading competence, should receive more attention in translator training and should by no means be taken for granted.