Speaker
Description
The recent development of neural machine translation has raised new questions regarding human-machine tandem translation, with machine translation post-editing (MTPE) being one of the key issues. To address the gap created by the lack of process-oriented post-editing (PE) observations via neural machine translation systems, the current project explored post-editing processes undertaken by translation trainees using the cloud-based computer-assisted translation platform Termsoup. Employing a mixed quantitative and qualitative approach, the project examined post-editing processes through Shih's (2021) three-tier model where PE changes result from both text and communicative functions.
Ten translation trainees were assigned 12 post-editing tasks spanning three distinct text types (informative, expressive, operative) in two language directions (Ch-En & En-Ch) in an undergraduate translation course in Taiwan. The analysis focused on the trainees' post-editing changes and errors, particularly in relation to the dominant communicative elements/functions in different text types.The PE changes and errors are further interpreted in conjunction with the trainees' PE logs, interviews, and data derived from the average PE time and number for each translation segment.
Preliminary findings indicate that trainees demonstrated the ability to identify varied communicative functions across different text types, employing diverse strategies to tackle the tasks. However, not all PE changes were deemed appropriate or accurate. An examination of PE errors, efforts, and student self-reports reveals that trainees tended to rely more on machine translation when tasks were perceived as 'difficult to translate.' This is also the context where errors in machine translations might persist. Meanwhile, trainees displayed mixed attitudes towards machine-assisted translation. Some praised how MT-assisted translation reduced cognitive demands in deciphering source texts and provided lexical or grammatical support. Conversely, some argued that post-editing could be more cognitively taxing than translating from scratch.
| Keywords | Pedagogical post-editing, post-editing processes, a communicative model of post-editing |
|---|
