Neural machine translation and the indivisibility of culture and language

Abstract

Neural machine translation (NMT) is becoming a common resource both for professional translators and for people who need small, occasional translations. The wide use of NMT for professional purposes reshapes the conditions of translation assignments. Moreover, the effects on the target text and language are not well known, although some studies already suggest a stronger influence of the source language on them. On the other hand, since many NMT tools are easily and freely available and even accessible via mobile devices, they are being increasingly used by non-professionals to carry out short translations mostly. These occasional translators use these tools both to understand a text originally written in a language they do not fluently speak (translation into their first language) and to publish a text in the language they do not master (translation into their second language or even into a language in which the user has poor or no knowledge at all). Finally, NMT is also present in everyday digital products even without users being aware of its intervention, for example in specific apps on smart devices. This article proposes a reflection on the effects of NMT in all these scenarios, with a special focus on the effects on the reception of the target text and on the target language standard.