Αξιολόγηση μηχανικά μεταφρασμένων νομικών κανονιστικών κειμένων στο ζεύγος αγγλικά-ελληνικά
Date Issued
May 2, 2023
Type
Μεταπτυχιακή Διπλωματική Εργασία
Abstract
Neural machine translation (NMT) systems have been gaining ground in recent years thanks to the improvement of the MT output, mainly at the level of fluency, especially in resource-rich languages. The increasing use of neural machine translation followed by post-editing by professional translators is directly related to the need for faster delivery times and at the same time lower costs, while there is still a need to maintain high quality. In the case of legal texts, in particular, the high level of accuracy and quality required still pose a major challenge to machine translation systems due to the specific features of these text types. Thus, the quality of the MT output may vary depending on the language pair and the legal genre. Using MQM-DQF error typology, this paper seeks to evaluate the quality of a) two machine-translated regulatory texts via eTranslation, and b) the post-edited translations produced by four professional legal translators in the English-Greek language pair, which remains under-researched. At the same time, the time taken by the translators for post-editing is recorded, as well as their attitude towards machine translation and post-editing of legal texts. Although the sample is small, the results show – based on the type and extent of errors found – that eTranslation does not yet lead to a satisfactory output in the said text types that have particular requirements due to the different legal systems, and as such, a significant amount of post-editing is required. Furthermore, the findings indicate that the final post-edited versions still contain errors which were either not found at all during the post-editing process, or were introduced while correcting the machine translation output, or the changes made were not applied consistently throughout the text. It is also interesting to note that each translator’s profile is directly associated with the edits made and, finally, it is necessary to highlight the importance of training in machine translation post-editing (MTPE) as well as the need to review the guidelines for full post-editing.
Subjects