A computer-assisted consecutive interpreting workflow: training and evaluation

Following a preliminary study that examined the potential effectiveness of a computer-assisted consecutive interpreting (CACI) mode, this paper presents a further trial of the CACI workflow. The workflow involves respeaking using speech recognition (SR) in phase I and production assisted by the SR text and its machine translation (MT) output in phase II. This study introduces a training framework for CACI that encompasses respeaking, sight translation, and post-editing. Additionally, it seeks to evaluate the CACI workflow with a group of trained students. Comparative analyses were conducted between conventional consecutive interpreting (CI) and CACI. Most of the findings from the preliminary study were successfully replicated in this study. The investigation revealed that CACI outperformed conventional CI in overall interpreting quality and accuracy in both directions of interpreting. Moreover, CACI exhibited higher fluency and lower cognitive load compared to conventional CI, albeit only in the L1–L2 direction. The quality of respeaking was found to be positively correlated with interpreting quality in both directions, underscoring the critical role of respeaking within the CACI workflow.