Validation of the Reliability of Machine Translation for a Medical Article From Japanese to English Using DeepL Translator

Abstract

Background: The reliability of DeepL Translator (DeepL GmbH, Cologne, Germany) for the translation for medical articles has not been verified yet. In this study, we investigated the accuracy of machine translation from Japanese to English for a medical article using the DeepL Translator. Methodology The subject of this study was an English-language medical article translated from Japanese, which had already been published. The original Japanese manuscript was translated into English using DeepL Translator. The translated English article was then back-translated into Japanese by three researchers. In turn, three other researchers compared the back-translated Japanese sentences with the original Japanese manuscript and calculated the percentage of sentences that retained the intended meaning. Results The mean ± standard deviation of the match rate for the entire article was 94.0 ± 2.9%. The match rate in the Results section was significantly higher than that in the other sections; while the match rate in the Materials and Methods section was significantly lower than the rate in the other sections. Compound sentences and sentences with an unclear subject and predicate appeared to be significant predictors for mismatched translation. Conclusions The translation for a medical article from Japanese to English was performed accurately by DeepL Translator.