In addition to Sulyok’s (2023) survey on revision and post-editing competence discussed above (1.5.6 Survey studies on PE competence and skills), a few empirical studies on post-editing have been carried out in recent years in a Hungarian context. They will be summarised and reviewed below.

Andrea Götz (2022) studied the target texts of 26 translation trainees, who post-edited the DeepL translation of a text on machine translation and post-editing. Götz determined in advance which parts of the MT output required post-editing, then she compared the changes introduced by students to this blueprint. She found that students did not perform most of the edits that would have been required and that most of their modifications were unnecessary and involved inserting synonyms. Interestingly, the less previous experience students had with post-editing, the more modifications they made, but most of these modifications were unnecessary, or they introduced mistakes in the text (Götz, 2022).

Kóbor and Szlávik (2024) investigated what effect a post-editing course had on students’ ability to detect errors in the MT output. 15 students from 2 Hungarian universities participated in the study, and they were asked to find errors in Google Translate-produced translations of journal articles at the beginning and at the end of the semester. Two different texts were used at the two data collection waves, and no control groups were involved in the study. The most important result of the study was that students’ capacity to identify errors clearly improved by the end of the semester (Kóbor & Szlávik, 2024). Perhaps even more interesting is the outcome of the qualitative analysis, which shows that the most frequently overlooked error type was what we could call “fluent mistranslation” in English. These were deceptively fluent phrases or segments that embodied grave mistranslations. Moreover, the authors found that students were exceptionally tolerant of errors in the titles. The final conclusion was that post-editing must be taught for several semesters and in several classes and that training error correction should be preceded by a relatively long period of learning to recognise errors in MT outputs (Kóbor & Szlávik, 2024).

Despite its novelty and strengths, the study is clearly limited by a small sample size and by the use of two different texts at the two data collection points. Furthermore, as there was no control group in the study, it is not clear what proportion of the improvement is attributable to the post-editing class itself and what proportion is the result of participating in translation training in general.

To date, the only monograph on machine translation and post-editing published in Hungarian was authored by Réka Eszenyi (2023), who not only gave an overview of the most important theoretical issues but also presented the results of her questionnaire and interview studies.

Chapter 5 in Eszenyi’s book presents a survey conducted with 30 translators from the Hungarian Department of the DGT. The aim of the survey was to discover translators’ attitudes to MT. Eszenyi arrived at the conclusion that translators think that MT brought positive changes to their work (it has become faster and more efficient), and that the job has not become more stressful as a result of technological development. As for creativity, Eszenyi’s data do not fully support the claim that translation has lost its creative nature. Finally, at the time of the data collection, translators did not perceive NMT as a threat to their profession (Eszenyi, 2023).

While the study provides valuable insights into Hungarian DGT employees’ attitudes to MT and PE, some methodological shortcomings should be highlighted here. Eszenyi’s conclusions are based on comparing mean values to the interpolated midpoint (2.50 – as she used a 4-point scale), but no one-sample t-tests were carried out to check whether mean values significantly differ from the midpoint. As these differences are small in most cases, it is hard to tell whether the conclusions are valid or rely on measurement errors. Nevertheless, the research represents the only attitude study in Hungarian post-editing research as yet and, as the author herself suggests, offers several opportunities for further analysis and perhaps re-analysis.

The other study Eszenyi (2023) presents is even more unique, as research on translators’ emotions towards their tasks, jobs and workplace is not particularly common in international research either, not to mention the Hungarian setting. Eszenyi carried out semi-structured interviews with seven professional translators and drew their profiles by identifying the dominant themes that emerged in the interviews. These motives are summarised in a table (Eszenyi, 2023, 171) and visualised on a mind map (Eszenyi, 2023, 173). The most common motives identified by Eszenyi include quality (of MT and PE), comparing and contrasting HT and MT, speed/temporal efficiency, pricing, digital joy/flow, and anxiety about the disappearance of the translator (i.e., of the profession). The profiles are even more interesting than the list of topics, and they provide a vivid snapshot of how Hungarian translators think and feel about MT and PE at the beginning of the 2020s.

After reviewing empirical research on post-editing competence in Hungary, we will turn to presenting the system of translator training in Hungary and in Szeged, as they form the social context of our research project.