The evaluation and reception of the translation quality of three translation modalities in live-streaming contexts: computer-assisted simultaneous interpreting, machine translation (MT) with human revision and raw MT

Abstract

This paper compares a complicated trio of testing the translation quality of different translational modalities in live-streaming contexts. These are computer-assisted simultaneous interpreting (CASI), machine translation with human revision (MTHR) and machine translation (MT). The data for the present study included 80 translation-majored undergraduate and postgraduate participants and their responses to a questionnaire and post-experiment interview. The investigation revealed that, based on participants’ self-perception, MTHR outperforms CASI mostly in accuracy, efficiency and clarity in thematically diverse live-streaming contexts ranging from chat shows, sports news and academic talks. However, informed by the emotion-driven NTRE model, which is adapted from Romero-Fresco’s (2017) NTR model, their self-appraisals differ, with CASI being slightly better than MTHR. Moreover, MTHR largely facilitates CASI, mainly supported by intralingual subtitles, although it maintains imperfections, mainly including punctuation issues, omissions, mistranslation and inappropriate collocation. This paper suggests that the evaluation and reception of live-streaming translations are text-type-specific and translation-mode diverse, offering critical insights into the theoretical, methodological and pedagogical contributions of novel practices, such as machine-assisted simultaneous interpreting and MT with human revision.