Translating highly specialised medical texts : analysis of outputs by a human translator and a machine

Abstract

Machine translation has a long and colourful history, and after more than seven decades of evolution in the translation of natural languages, great advancements have been made in the field, especially in recent years with the emergence of neural machine translation which has become the mainstream approach to machine translation. As a result, machine translation is gradually finding its way into specialised domains. This study sets to present two outputs: one produced by a human translator and another produced by DeepL, a neural machine translation tool. This was performed with the view of: testing the ability of neural machine translation softwares with regard to producing an adequate translation of a highly technical medical text, i.e. an excerpt extracted from Anatomie des Centres Nerveux, a 19th -century anatomical work; determining the error that dominates neural machine translation outputs; comparing the machine translation with the human translation to test for likeness and differences. The findings of this study demonstrate that regardless of the great advancements in the field of machine translation, machine translation softwares are not yet able to deliver adequate and hence, accurate, reliable, and comprehensible medical translations without human involvement.