What is the present MT-development status? To what extent are human translators still required?

  • Published: 3 ans ago on 17 janvier 2018


  1. Aljoscha Burchardt dit :

    Why do people still use their kitchen when there’s Mc Donald’s? To be serious, today’s MT quality is not reliable enough to show its results to paying customers without human intervention. On the web, we see a lot of instant translations, e.g., on eBay or Facebook, but these are use cases in which nobody would have payed a translator anyway.

    Sooner or later we will probably see machine output that is « good enough for the job ». I know that translators are allergic to this statement and I was part of many lively discussions on this. As a trained computational LINGUIST, I am a language-lover myself and I rather not read a book than reading a bad translations. But I am afraid that there is a simple logic of business we cannot change. Often translators argue that we need to educate the customer to value a good translation, but I am afraid that it will not work this way.

    I am nevertheless convinced that translators – like many other people who’s jobs have changed due to digital transformation – will have a lot of work to do in the future. But the tooling and working environment will change. Those who are open to this and adapt to new possibilities (or even participate in the development) will most probably be able to offer good language services that will make the customers happy.

  2. Arle Lommel dit :

    I agree with your points, Aljoscha. I might add that all this MT requires a lot of human translation to train and stay current. It doesn’t come from nowhere. The ability to reuse human translation in this fashion actually makes it a more valuable resource (although it does increase competition for low-value translation work).

    The expectation the content will be translated means a bigger pie for translators, not a smaller piece of a pie that stays the same size. Research I’ve done shows that demand for human translation will continue to increase for the foreseeable future, even as demand for MT increases, so MT is *not* replacing human translation, but instead supplementing it.

    That isn’t to say that there aren’t jobs and project that will be lost to MT, but the net additions outweigh those losses. And the human translation jobs that will be replaced by MT are the ones that were marginal to begin with. The ones that are being added will be much more solid.

    I compare translators to accountants. Today a lot of what accountants used to do (the bookkeeping aspect) has been automated. But that didn’t mean that we don’t have accountants any more. They actually make better money now than thirty years ago, and they focus on things like predicting business issues, managing business flows, and anticipating problems before they occur.

    I expect that translators will see a similar sort of shift as they shift to focus on interesting and difficult things that require their knowledge and sensitivity in a way that translating the five hundredth variant of “click ‘OK’” never will.

    I like the McDonald’s analogy, by the way. Even as prepared foods have become available, grocery stores haven’t disappeared and sales of many hand-crafted ingredients have gone up. Your home kitchen, a fast-food restaurant, and a nice restaurant all serve different purposes, and none can replace the other.