Abstract

Machine translation has become more popular in recent years and is now widely used by a broader public, such as laypeople, professional translators or those who need to produce academic texts in a language they do not master (Bowker & Buitrago Ciro, 2019). Even with the advances in technology, it is still necessary to proofread (or post-edit) machine-translated texts. However, this should not require more effort than producing texts directly in the target language. This study aimed to analyze the length of time to complete three different tasks involving L1 Brazilian Portuguese and L2 English: writing in the first language (EL1), writing in the additional language (EL2) and post-editing in the additional language (PEL2). The processes of these textual productions carried out by three PhD candidates in the field of education (P01, P02 and P03), who performed these tasks in different orders, were analyzed. Data was collected with a prospective questionnaire, Translog©-II program, Tobii Studio Eye Tracking Software©, OBS Studio, a smartwatch, and guided retrospective protocols. Results have shown that EL2 required more time from the participants, possibly due to their intermediate L2 proficiency level. Nevertheless, the time spent on this task was inversely proportional to its position in the order of the tasks. It has also been observed that machine translation has a facilitating effect on PEL2, requiring less time than EL2. Finally, it has been noticed that the temporal effort of EL1 and PEL2 altogether was inferior to that of EL2. Although post-editing is not performed very often in academia and requires more effort for more complex texts, results have suggested that it may be a more efficient form of writing in L2 when writing in L1 and post-editing in L2 require less time than writing in L2.

Keywords
academic writing; machine translation; post-editing; temporal effort