The truth about machine translation today

The truth about machine translation today

“Automatic translation of all texts in all the world’s languages and automatic interpreting of spoken speech is here! The language barriers will come down within ten years!” Articles making such claims have recently been seen in such prominent publications as the Wall Street Journal, the International Business Times and Forbes. It all sounds like science fiction – something straight out of Star Wars, The Hitchhiker’s Guide to the Galaxy or Star Trek.

Published 2/1/2017

Not too many years ago, such assertions would have been dismissed as nonsense. It worries me to say that these claims are now viewed as highly believable, and even as accepted truths.

Promising the earth

Machine translation promises instant free translation. These three words are enough to get anyone who runs an international business salivating. The truth is that translation and interpreting have long been seen as problematic and expensive necessary evils that crop up when all you want to do is deliver a product. They are seen as unnecessarily troublesome issues that should ideally be solvable without having to spend too much time and money. So it’s not surprising that when offered a free technical solution with the shortest lead times, some people get a little over excited and tend not to take too critical an approach.

The reality is, of course, a little less perfect and wonderful. Unfortunately, the many article writers and technophiles (myself included) who preach about the benefits of the new technology have failed to convey the circumstances under which machine translation works best and the situations where there are still a number of problems and drawbacks. As a result, the businesses that are attracted to machine translation sometimes have blind faith in what it can deliver.

The importance of realistic expectations

A number of unrealistic expectations about machine translation crop up time and time again. I believe that taking a closer look at these could save us all – buyers and suppliers alike – from future problems and disappointments. Here are a few common inaccurate expectations:

  1. Machine translation is free. In some cases, it can be. Take Google Translate, for example. However, businesslike, professional machine translation isn’t free – although it can be much cheaper than paying a human translator. A language technician still needs to build and maintain the translation engine, and to be paid for the work involved. A really good machine-translated text will also have been quality assured by humans. So machine translation does still involve a cost.
  2. Machine translation produces good translations. Well, that’s almost true, but of course it depends on many different factors. Machine translation usually produces good translations, but it’s important to understand that you can never fully rely on the accuracy of a text that has been machine translated without human involvement. You should also remember that what ‘good’ means varies for different texts, target groups and situations. You also have to weigh up the risks against the opportunities. The immediate risks associated with machine translations of work e-mails about business strategies pale in comparison with the risks associated with machine translating the instructions for a defibrillator.
  3. Machine translation will make human translators completely redundant. Absolutely incorrect. See the previous point. Human translators will always be needed. However, using machine translation will enable us to translate more material, which is a positive thing in view of globalisation.
  4. Everything can be translated using machine translation. The statement itself might be true, but you probably shouldn’t translate everything using machine translation without human involvement. Instead, you should be extremely selective about which texts are machine translated to get the right quality for the right type of text. However, machine translation can almost always be used as an additional tool alongside human translation.
  5. The language barriers will come down within ten years (or similar statements). IBM made this claim as early as 1955 when they built their first translation engines. With the benefit of hindsight, it might be worth taking such statements with a good pinch of salt.

The revolution is already here – long live machine translation!

Despite everything, there’s no doubt that machine translation has already revolutionised our industry and will continue to change it. If we extrapolate what today’s technology can do for us a few years into the future, we end up not far from the science fiction scenarios I mentioned at the beginning of this article. But we need to keep our expectations realistic. Automatic translation and interpreting is likely to be of better quality in the future, but will never be perfect in all respects. Large quantities of information will certainly be translated and interpreted using machines in future, but at the same time we will always need human interpreters and translators.

Semantix’s mission

Semantix’s mission is to contribute towards international business and a multicultural society. For us, making use of modern technology is clearly the key to achieving this. We will therefore continue to invest in technology and to develop it. In doing so, it is important that we help others not to trust blindly but to see with open eyes the realistic and wonderful potential of the emerging developments within translation technology.


Latest blogposts