A wake-up call at the Translating Europe Forum

The Directorate-General for Translation (DGT) of the European Commission staged its 5th Translation Europe Forum in Brussels on 8-9 November 2018. About 500 stakeholders from the translation community (language service providers, producers of translation tools, academics, researchers, university students, etc.) attended this event, which focused on ‘translation in the age of data’. 

Those of us who attended from the Translation Centre came back to Luxembourg with a fresh perspective on our changing world.

A large panel of artificial intelligence (AI) engineers, research scientists, university professors, and linguists and senior staff from the EU institutions, including Commissioner Oettinger, addressed the conference, offering a comprehensive overview of the position of translation today.

It was great that the DGT had also invited a philosopher and a graphic reporter to the conference. With their specialist presentations, they helped participants gain a broader perspective on the current ‘digital revolution’.

The philosopher, Karim Benammar, invited the audience to face the facts: the translation community is clearly in a transition period. The old world is gone but the new one is not yet ready: translators no longer translate a document on a blank sheet of paper from scratch but machine translation is still far from being a substitute for human translators.

This transition period was eloquently depicted by the graphic reporter:

As the picture clearly shows, the translator, represented as a cyclist, is journeying through a vast amount of data on a bicycle with square wheels. It doesn’t look very comfortable, does it?! In other words, there is a huge volume of data, but we do not yet have adequate tools to tap into it efficiently. In addition, there is still a lot of room for improvement in the ergonomics of translators’ working environments.

With a view to improving translators’ interfaces, a number of AI researchers are already focusing their attention on how to switch from ‘segment-level’ to ‘document-level’. This new approach offers numerous advantages: the metadata of the document to be translated can be used to clarify the context, pictures can help interpret the text (a ‘Chair’ surrounded by several people is unlikely to be translated as ‘seat’), the gender can be adapted through all the segments automatically, etc.

Things are clearly moving forward in the AI field. This is causing the whole translation community to question its working methods and to wonder what those who want to be translators in the future will have to do.

First of all, translators must adapt. It is no longer a question of human vs. machine translation (MT), but rather how we can work efficiently WITH the machines. It is not necessary for translators to become AI engineers but they will have to understand how the machines work, be aware of the technology’s limitations and learn how to make the machines more efficient (if possible by using coding). They will also need to make sure that their data (the ‘translator’s oil’, as it was repeatedly described) is clean. Learning how to post-edit machine-translated texts is also crucial.

Secondly, translators should position themselves to reflect what they are: language experts. They are not just ‘translators’ (like the eponymous software that just replaces words) but people with a range of specific skills (terminology management, translation memory maintenance, desktop publishing, accountancy, etc.). This was underlined by several speakers who argued that language experts should be paid hourly like lawyers, consultants, etc., instead of by the word.

Thirdly, language experts should ensure that their clients’ expectations of AI are managed properly. The media’s current loud and clear message is that MT is ready, but those in the language business know perfectly well that this is not the case. Language experts should clearly inform their clients what is feasible and what is not.

Although translation is clearly in a transition period, all conference participants agreed that there would be good times ahead for the simple reason that the current globalisation process requires an exponential number of language barriers to be overcome. Machine translation is proving to be very helpful in this context as it offers a great alternative to zero translation. However, human translators really must start to recognise and exploit the added value they bring to multilingual communication compared to machines, so as to safeguard their future.