Objective To assess the ability of Google Translate (GT) to accurately interpret single sentences and series of sentences commonly used in healthcare encounters from English to Spanish.
Design English-speaking volunteers used GT to interpret a list of 83 commonly used sentences and series of sentences of different lengths containing both medical and non-medical terminology. A certified medical interpreter evaluated whether the meaning of these sentences was preserved.
Participants Eighteen English-speaking subjects (nine males and nine females), with a mean age of 36 years, volunteered for this study to read sentences.
Main outcome measures The accuracy of GTs (1) real-time voice recognition (ie, transcription) of English sentences, (2) real-time translation of these transcribed English sentences to Spanish, and (3) GTs speech synthesis ability to preserve the meaning of spoken English sentences after translation to Spanish.
Results Speech synthesis accuracy, with preservation of the original English-spoken sentence(s), was 89.4% for single sentences with ≤8 words; 90.6% for single sentences with >8 words; 52.2% for two sentences and 26.6% for three sentences. Furthermore, the number of transcription and translation errors per sentence(s) significantly increased with the number of sentences (p<0.05).
Conclusions Despite the fact that GTs accuracy was widely variable and dependent on the length of the spoken sentence(s), GT is readily accessible, has no associated monetary costs, and offers nearly immediate interpretation services. As such, it has the potential to routinely facilitate effective one-way oral communication between English-speaking physicians and Spanish-speaking patients with limited English proficiency.
- assistive technology
- medical apps
Statistics from Altmetric.com
JB, HJ and RS contributed equally.
Contributors RS conceived of the presented idea. JB and HJ conceived and planned the experiments. JB, HJ, DC, MM, NM and SS carried out the experiments. JB and HJ collected and analysed the data. JB and HJ wrote the manuscript with support from all authors. HD, JC and HL supervised the project.
Funding RS was funded, in part, by NIH grant number T32 GM008620-17.
Competing interests None declared.
Patient consent for publication Not required.
Data availability statement Data are available on reasonable request. Deidentified participant data are available from researcher JB, whose ORCID identifier is 0000-0002-2133-7708. Reuse is not permitted unless previously discussed with researcher.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.