A SYSTEMATIC REVIEW OF STUDIES ON COGNITIVE EFFORT IN MACHINE TRANSLATION POST EDITING AND HUMAN TRANSLATION
Author :
Abstract
Keywords
Abstract
Machine translation (MT) has undergone numerous changes since it has been used as a strategic and political tool during the Cold War. While the first machine translation models were far from human translation (HT) quality, current models are showing promise. However, the post-editing (PE) of machine translation output is still needed and it demands varying levels of cognitive effort, which can influence productivity, translation quality and even job satisfaction. Technical and temporal aspects of post-editing have been more extensively studied, while cognitive effort remains relatively underexplored. This systematic review investigates cognitive effort in translation processes, focusing on HT and machine translation post-editing (MTPE). Studies were included if they empirically measured cognitive effort in MTPE and HT using eye-tracking, keystroke logging, pause analysis, or self-reported measures. Relevant literature was identified following PRISMA across Dergipark, HAL, ProQuest, ScienceDirect, Scopus, and Wiley Online Library. A total of 3482 records were identified, with the last search performed in June 2025. 52 studies met the inclusion criteria, comprising 30 on HT, 14 on MTPE, and 8 on a combination of HT and MTPE, utilizing methods such as eye-tracking, keystroke logging, and think-aloud protocols. Limitations include the small number of MTPE versus HT studies (n= 8). Our analysis revealed that MTPE generally requires less cognitive effort than HT. This difference is affected by factors such as participant expertise, text complexity, and MT output quality. However, phase-specific analyses indicate that certain stages of MTPE, especially initial orientation, can demand more cognitive resources than equivalent stages in HT.
