How People Living With Amyotrophic Lateral Sclerosis Use Personalized Automatic Speech Recognition Technology to Support Communication.

The study had only three dyads, and the sample was not balanced across demographic or other characteristics, so the findings are not representative. All participants were already aware of Relate and were using it or recording phrases before joining, which may select for people already interested in the app. The app was English only and Android only. Word error rate and meaning ratings do not fully capture listener understanding, because the person and partner could still use speech, gesture, context, and other cues, and captions were often not looked at. Custom cards require planning and at least 24 hours to update, but participants found it hard to predict which words would be used. Relate transcribes all voices, so conversations with overlapping speech or background talk were less suitable. Only Bob showed FDA-2 decline, so the study says little about how caption usefulness changes as speech declines for all people with ALS. The researcher was a specialist speech therapist who also collected and analysed the data, and meaning ratings were made by speech therapists rather than by the conversation partners themselves.

The supplied text lists research support from the National Institutes of Health, extramural. Ethical approval was granted by University College London Ethics Committee. No author conflicts of interest are stated in the supplied text.

Do not conclude that personalized ASR reliably improves communication for people with ALS. This was a three-dyad qualitative study, and participants found captions too inaccurate for many conversations, often using other AAC instead.