Machine translation in medicine : a neural machine translation comparison of French 19th century anatomical text

Abstract

Over the last decades, the translation industry has gone through unprecedented changes due to the advances in technology. The use of Machine Translation has grown exponentially and is slowly being introduced in specific-domains such as that of medicine. Consequently, this dissertation aims to investigate whether the neural machine translation applications Google Translate and eTranslation, are able to produce satisfactory translations in the field of human anatomy. A section from the 19th century French work ‘Anatomie des centres nerveux’ by the neurologists Joseph and Augusta Déjerine will be compared and through the process of post-editing and error analysis, the development, strengths, weaknesses and the outcome of machine translation will be highlighted. The theoretical part of this dissertation is concerned with neural machine translation and medical language whereas in the practical part, the researcher post-edits the raw MT output and generates a taxonomy of errors. The findings illustrate that despite the revolutionary developments in machine translation due to neural networks, machine translation systems are not yet suitable to translate the Déjerine’s anatomical text. The analysis depicts that the output quality is inferior but through post-editing a high-quality final product can be attained.