Márta Lesznyák: A longitudinal investigation of translation trainees’ translation and post-editing competence development I
›
Introduction

Introduction

Jegyzet elhelyezéséhez, kérjük, lépj be.!

As artificial intelligence is transforming the translation industry, the role of human translators is being redefined. These days, their work primarily includes post-editing and quality assessment, which presents a great dilemma for trainers: what should we teach – translation or post-editing – and at what stage of training? This volume presents the results of a longitudinal study examining the development of both traditional human translation competence and post-editing competence. The findings may help rethink translator or language professional training and provide some ideas on how to sequence and focus instruction.

Jegyzet elhelyezéséhez, kérjük, lépj be.!

It is unquestionably the emergence and the spread of Artificial Intelligence (AI) that have attracted the largest attention and caused the most significant transformation in recent years in both everyday life and academia, including translation studies and the translation industry. Although the public is preoccupied with large language models (LLMs) and tends to equate AI with LLMs, it should be stressed that AI has long been a part of the translation profession, as statistical and neural machine translation engines have already utilised artificial intelligence.

Jegyzet elhelyezéséhez, kérjük, lépj be.!

For morphologically rich languages like Hungarian, the real breakthrough in automated translation was the arrival of neural machine translation (NMT). Although the idea of NMT has long been known to IT specialists, it was not until the middle of the 2010s that technology and hardware became mature enough to cater for the needs of NMT systems. Google switched to NMT in 2016, and it slowly became clear to everyone that Hungarian lost its immunity to machine translation (MT). Previous MT systems did not produce accurate and well-formed translations when Hungarian was involved but it changed with the advent of NMT. By 2020, it was inevitable that MT and post-editing would form part of the future for translators, whether we like it or not.

Jegyzet elhelyezéséhez, kérjük, lépj be.!

To address this need, the Translation and MT Post-Editing Competence Research Centre was set up at the University of Szeged in March 2020 with the purpose of studying the elements of translation and post-editing competence. A comprehensive research project began in May 2020, which involved a longitudinal component. This volume presents the results of the longitudinal part of the project, focusing on how translation and post-editing competence develop when students work with a text whose form and content are relatively familiar to them. The analysis also considers the role of different subcompetencies and certain background factors in translation performance. A subsequent monograph will address the longitudinal data on students’ development when working with relatively unfamiliar and more challenging text types, specifically legal texts.

Jegyzet elhelyezéséhez, kérjük, lépj be.!

Initially, 31 students entered the “bilingual” longitudinal investigation presented in this paper, but because of data loss, certain aspects of the analysis are based on reduced datasets. The analysis includes a comparison of students’ translation and post-editing performance measures (i.e., the quality of the target text (TT) and time on task) at the beginning and at the end of their studies. In addition, changes in some competence elements and other background factors are also examined. These elements and factors include language competence, thematic knowledge, beliefs about translation and task perception. Finally, changes in correlation patterns are analysed, too. Throughout the study, the translator and the post-editor group is compared.

Jegyzet elhelyezéséhez, kérjük, lépj be.!

The main findings indicate that there is a considerable development of both translation and post-editing competence. By the end of the training, both the HT and the PE groups worked faster and committed fewer errors. Nevertheless, the PE group outperformed the HT group both at the beginning and at the end of the training. Moreover, their improvement was more pronounced, in spite of not receiving targeted training in post-editing. These findings suggest that post-editing could indeed be a more efficient work form than traditional from-scratch translation. More importantly, they indicate that post-editing skills can develop through traditional translation training, too.

Jegyzet elhelyezéséhez, kérjük, lépj be.!

At the same time, it is evident from the data that a targeted intervention to enhance post-editing skills would be very much needed to boost development. It is also worth noting that the two groups show a slightly different developmental path: the HT group demonstrated the greatest improvement in accuracy, while the PE group improved primarily in fluency.

Jegyzet elhelyezéséhez, kérjük, lépj be.!

The study’s findings have implications primarily for training. While human translation may have lost much of its relevance as a standalone professional activity, it appears to serve a valuable pedagogical function by preparing the ground for post-editing. Just as it is important for children to learn manual mathematical calculations before relying on calculators or computers, it is equally important for future language professionals to engage in from-scratch translation. This practice helps them develop the ability to recognise possible flaws in machine-translated texts. In addition, training in translation contributes to developing post-editing competence.

Jegyzet elhelyezéséhez, kérjük, lépj be.!

Furthermore, our findings underscore the need to raise students’ awareness of the limitations of machine translation. Students must learn decision-making skills (i.e., how to decide whether the MT output is correct) and special attention must be paid to terminology, which appears to be one of the less developed sub-skills of student translators. As humans will still be needed for some time in the translation, post-editing and quality assurance processes, it is vital that students receive proper training in post-editing and in its sub-skills listed above. The findings of the present investigation offer insights into the developmental process, thereby possibly guiding and informing training strategies.

Jegyzet elhelyezéséhez, kérjük, lépj be.!

This volume is organised as follows. In Chapter 1, some basic terms are defined. This is followed by the presentation of translation competence models and translation competence development models. In addition, research related to translation competence development is summarised. The next chapter focuses on post-editing competence and its development. Then, research on post-editing competence is discussed and research on translation and post-editing competence in Hungary is described. As a summary of the theoretical overview, the lack of longitudinal research on post-editing competence is highlighted. To give context to the research, translator training in Hungary and in Szeged is portrayed.

Jegyzet elhelyezéséhez, kérjük, lépj be.!

The methodology chapter begins with the description of the comprehensive Szeged competence research project. Next, the research questions of the present study are listed and the methodology of the longitudinal part of the project is described in detail. This is followed by the presentation of the results and their discussion. The paper ends with a summary that includes suggestions for translator training and further research.

Jegyzet elhelyezéséhez, kérjük, lépj be.!

Finally, it should be noted that both DeepL Write and ChatGPT-4.0 were employed to support the proofreading and editing of some sections of the English text, particularly the Introduction and the Conclusions. Suggestions generated by DeepL and ChatGPT were always carefully evaluated and frequently rejected. In many cases, a “dialogue” was initiated with ChatGPT, in which the AI was asked to assess human(author)-revised versions of ChatGPT-suggested editions.

Jegyzet elhelyezéséhez, kérjük, lépj be.!

Unlike ChatGPT, DeepL Write does not require a prompt, but it offers configurable settings for register, and Academic style was consistently selected while using it.

Jegyzet elhelyezéséhez, kérjük, lépj be.!

As for ChatGPT the following prompts (and slight variations were used):
 

Jegyzet elhelyezéséhez, kérjük, lépj be.!

Proofread the following <sentence/paragraph>. Edit it using academic English. Introduce only necessary changes. Make a list of your changes and explain/justify them. Polish the following <sentence/paragraph>. Use academic English. Proofread the following <sentence/paragraph>. Does it sound correct academic English? Suggest alternatives for <word/phrase> in academic English. Please proofread the following text for grammar, clarity, and academic style. Use formal, natural-sounding English suitable for publication. If needed, split long sentences, enhance coherence, and improve word choice. Do not simplify technical content.
 

Jegyzet elhelyezéséhez, kérjük, lépj be.!

ChatGPT was also asked to create the ideal prompt for itself. It looks as follows:
 

Jegyzet elhelyezéséhez, kérjük, lépj be.!

Please proofread the following text for grammar, clarity, and academic style. Use formal, natural-sounding English suitable for publication. If needed, split long sentences, enhance coherence, and improve word choice. Do not simplify technical content.
chevron_left
chevron_right
Tartalomjegyzék navigate_next
Keresés a kiadványban navigate_next
A kereséshez, kérjük, lépj be!
Könyvjelzőim navigate_next
A könyvjelzők használatához
be kell jelentkezned.
Jegyzeteim navigate_next
Jegyzetek létrehozásához
be kell jelentkezned.
    Kiemeléseim navigate_next
    Mutasd a szövegben:
    Szűrés:
    Kiemelések létrehozásához
    MeRSZ+ előfizetés szükséges.
      Útmutató elindítása
      delete
      Kivonat
      fullscreenclose
      printsave