Writing for machine translation
Machine translation has improved significantly in recent years and is becoming more and more widespread. Documents drafted in a multilingual environment therefore stand a good chance of being translated by a machine. Authors should be made aware of this and be able to access tips about how to write ‘machine-friendly’ texts. The Translation Centre has just released a booklet on ‘Writing for machine translation’ now available on its website.
As the Translation Centre uses neural machine translation for some of its services, its linguists and language technology experts have prepared some guidelines for authors based on their own experience. As they have observed, the source texts themselves are central for the quality of the machine translation produced.
When writing for machine translation, authors should first of all follow the principles of clear writing, which apply to any document. In addition, they should be aware of pitfalls that can ‘confuse’ the machine translation tool. These can be related to language, content and layout. The short guide ‘Writing for machine translation’ provides some real-life examples in various EU languages that have been taken from CdT translations. It shows what happens if sentences are too long or too short, terminology is inconsistent or punctuation is used wrongly, for example.
As the Translation Centre’s expertise in the machine translation field grows along with new technological advances, the guide on ‘Writing for machine translation’ will be subject to review. Machine translation will continue to grow, as it offers many advantages including the fact that it expands the ways in which organisations can leverage multilingualism and reach much larger audiences than before.