Real-time voice-to-voice translation sometimes failed to recognise words due to background noise or dialects
Source
Panayiotou (2020). The perceptions of translation apps for everyday health care in healthcare workers and older people: A multi-method study. Journal of Clinical Nursing.
Description #
Staff reported that not all experiences with translation apps were positive: real-time voice-to-voice translation using apps such as Google Translate sometimes failed to recognise spoken words accurately because of background noise or the different dialects spoken by patients, at times producing nonsense output.
"There were instances in which real-time voice-to-voice translation using apps such as Google Translate™ had not recognised the words accurately due to background noise or different dialects spoken by patients" (Panayiotou, 2020, p. 3521)
"Sometimes it brings up gibberish." (Panayiotou, 2020, p. 3521, participant from healthcare staff consultation)
Methods Context #
What? #ⓘ
The observable: the accuracy and functioning of real-time voice-to-voice translation as experienced by staff during use and appraisal.
"Have you ever used technology to translate information from one language to another before?" (Panayiotou, 2020, p. 3518)
How? #ⓘ
Focus group discussion of experiences with translation apps plus hands-on appraisal of A-0016ArtifactA-0016Initial AI draftGoogle TranslateA free, general-purpose machine translation app increasingly used as an ad-hoc communication tool in healthcare settings to bridge language barriers with LEP patients, including via voice-to-voice translation. "One such…'s voice-to-voice feature, analysed by inductive content analysis.
"Google Translate™ allowed for real-time audio or text translation and required an internet connection." (Panayiotou, 2020, p. 3518)
Who? #ⓘ
Nursing and allied health staff (n = 17) from three Melbourne metropolitan public health services.
"Seventeen healthcare staff from three Melbourne metropolitan public health services took part in the consultations." (Panayiotou, 2020, p. 3519)
Other Notes #
Qualitative failure-mode finding; no associated figure/table (text-only). Supports the preference for fixed-phrase over voice-to-voice apps.