Evaluating the accuracy of speech to text applications for cochlear implant candidates during COVID-19.
Cochlear Implants Int
; 24(1): 1-5, 2023 Jan.
Article
em En
| MEDLINE
| ID: mdl-36148962
ABSTRACT
OBJECTIVES:
Universal mask wearing due to COVID-19 has introduced barriers to clear communication. In hearing impaired individuals this can impact informed surgical consent. For cochlear implant candidates, who do not rely on sign language, real-time transcription with a stenographer (CART) is the gold-standard in assistive technologies. If CART is not available, speech to text (STT) applications have been advertised as solutions, but their transcription accuracy with or without an N95 mask is not well-established. Herein, we sought to investigate the transcription accuracy of three STT solutions for iPhone and compare their performance to the CART service at our institution.METHODS:
Three native English speakers and three non-native English speakers read two passages (a cochlear implant consent and the non-medical 'Rainbow passage') with and without an N95 mask. Error rates from the comparison of the transcript (from either the STT app or CART) with the original passage were calculated.RESULTS:
The CART service had the lowest error rate of all testing conditions (4.79-7.14%). Ava 24/7 (15.0 ± 9.49%) and the iPhone dictation (15.6 ± 6.65%) had significantly lower average error rates than the Live Transcribe (37.7 ± 20.3%) (P < 0.0001) application. Neither the presence of an N95 nor the type of passage had a statistically significant impact on the error rate.CONCLUSION:
CART should be used to augment communication with patients who are hard of hearing. If CART is not available, a STT application such as Ava 24/7 or the native iPhone dictation application may be considered, even in the context of medical terminology.Palavras-chave
Texto completo:
1
Coleções:
01-internacional
Base de dados:
MEDLINE
Assunto principal:
Percepção da Fala
/
Implantes Cocleares
/
Implante Coclear
/
COVID-19
Limite:
Humans
Idioma:
En
Revista:
Cochlear Implants Int
Assunto da revista:
AUDIOLOGIA
Ano de publicação:
2023
Tipo de documento:
Article
País de afiliação:
Estados Unidos