OR How MT is wangling its way into literature

With the resistance against the use of MT that I am experiencing, on the one hand by many of our inhouse translators and on the other by some of our clients, I had been assuming that such resistance would be even stronger when it comes to literary translation. I was wrong. I had also assumed that in academia and at universities where translation is being taught, MT would be a taboo area. Wrong again.

I was fortunate enough this weekend to find myself on the website of the Goethe Institute (an important cultural institution of the Federal Rep. of Germany that promotes cultural exchange, education and literature; they are particularly active in promoting literary translation between Germany and England). Much to my surprise their webpages are packed with really fascinating articles and talks about the subject.

One of these articles that I would recommend to anyone in our business is “Post-human Literary Translation?” by Duncan Large, Academic Director of the British Centre for Literary Translation at the University of East Anglia in Norwich: https://www.goethe.de/ins/gb/en/kul/lue/ail/21984887.html

His focus is on how the rapid development of neural machine translation is starting to convince literary translation academics and (to a lesser extent, literary translators themselves) that resistance against MT is no longer sensible. He claims that it is most likely that the future of literary translators – just like that of their “commercial” or “technical” colleagues – is going to be shifting towards post-editing.

Prof. Large uses the introductory sentence of Kafka’s The Metamorphosis (“Als Gregor Samsa eines Morgens aus unruhigen Träumen erwachte, fand er sich in seinem Bett zu einem ungeheueren Ungeziefer verwandelt.“, comparing the output of a number of MT systems – and also of various human translations. In particular the “ungeheures Ungeziefer” has been rendered in many different ways. Surprisingly (or perhaps not), the machine systems provide a rather smaller variety of solutions than their human counterparts. Similarly, “unruhig” (of sleep) has been translated as “troubled”, “agitated”, “anxious”, “fitful” by various human translators, each emphasizing a different interpretation or nuance. The machines on the other hand offer more convergent translations: Their solutions tend to be on the “safe” side.

The conclusion of his article is that MT still has a long way to go when it comes to literary translation, because clearly, the engines lack knowledge of the real world and cannot “interpret” meaning as such – and what they lack mostly is human creativity. And let’s not forget that we as readers expect much more of literary translations than of commercial ones, so we tend to put the benchmark much higher.

It has been my belief for quite a while now that we should not ask MT to be perfect and produce pieces of art. That is simply not what they are meant to do. But these perceived failings should not prevent us from making use of MT and accepting it as a great aid to our own efforts. The trend is definitely towards MT, refined by humans.

I believe that we also need to be honest, and humble when assessing MT vs. HT. We often make the mistake of ridiculing one particular sentence that DeepL has produced by comparing it with our superior human version. But if we ask ourselves: How long did it take a human translator plus a reviewer to come up with that superior version we have to admit that we may have gone through 4 or 5 iterations until we came up with that perfect version. And we all know how relevant the time factor is in our business. If a client wants a press release of 800 words within an hour – how many of us would be able to produce an error-free, fluent, attractive translation in that time if we were left to our own devices? So, is it not a perfectly sensible idea to let MT help us on our way?

Let me finish with a quote from Prof. Duncan:

“The role of human literary translator in its current form seems set to be largely phased out, but that doesn’t automatically mean mass unemployment for translators… Human translators will continue to enjoy the process as well as the product, and there will be readers who enjoy the human-translated product, too – which will command premium prices, like line-caught fish in the age of the factory trawler, or a hand-built Aston Martin in the age of the production-line robot.”

Meanwhile, those of us who are not producing literary translations but more mundane ones, let’s focus on the task at hand and turn ourselves into excellent post-editors. Because that is what is needed and what the market will pay for.

If you are unsure about your aptitude as a Post-Editor I would recommend going to: https://www.goethe.de/ins/gb/en/kul/lue/ail.html and scrolling down to the QUIZ towards the bottom of the page. See how well you do in identifying translations by Humans or AI. If you score 10 you have what it takes to be a talented Post Editor.