This month has seen a series of claims from Google which make interesting reading for translation and localisation professionals. For example:

Google translates more text in a day than all the world’s human translators can do in a year.

With 200 million current users, Google translates enough text every day to fill a million books.

Perhaps it does. But would you really want to read those books?

A friend and client of mine recently directed me to this picture of a novelty coffee table that was made from the remains of a crashed Ferrari. The table reminds me of a machine translation of the original car. You can see the licence plate, you can see the badge, and you can tell that this used to be a Ferrari. But you certainly can’t drive it.

Machine translation is a reality, and as the machines grow more sophisticated it’s likely that, in time, the role of the human translator will develop into something more editorial.  But the requirement for some level of human involvement will surely remain.

A recent report on attempts by the United States armed forces to develop a machine translation device as a substitute for human interpreters in Afghanistan illustrated the limitations of even the most advanced non-human translation.  The Defense Advanced Research Projects Agency (DARPA) funded a five years research project.   DARPA’s distinguished fifty year track record has seen it light the way for some of the world’s most complex and high profile technical research projects. The agency has a $3.2 billion annual budget, and a sizeable chunk of that money was directed towards machine translation research. Five years of tireless effort and bottomless funding failed to produce a software solution that even came close to replicating the “human factor”.   DARPA’s speech-to-speech translation system eventually achieved an 80% accuracy rate, but attempts to use the final product in place of human interpreters ended in failure and, at worst, farce.

Conveying the sense and spirit of a message fundamentally requires a human intelligence and sensibility.  One Wolfestone partner who understands that very well is Debbie Bouwer, owner of The WordWright.

Debbie’s expert copywriting service ensures that the sense and spirit of a client’s message is conveyed in the most persuasive way possible. Significantly, The WordWright offers a localisation service that takes into account the subtle differences between the English spoken in different Anglophone countries.  The variations between the language spoken and written in, for example, South Africa, Australia, the United States and the UK are notable enough to cause serious communication issues for those who disregard them.  The WordWright has built an enviable reputation by paying attention to detail where it matters most.  This begs the question, when clients can benefit so clearly from the “human touch” of a localising copywriter, why would anyone compromise their message by attempting to convey it from one language to another without any human contribution at all?

Fifteen good reasons not to machine-translate a restaurant menu

Here is a selection of familiar restaurant dishes which have been machine translated into a new language. These are real life examples taken from actual menus. Anyone wishing to guess the original dishes is welcome to log a comment:

Duck denies child support sour

Irrational beef pot

Unconscionable lettuce salad fries the wheat field

Chicken demands cabbage lifestyle

Renegotiated vegetable strand pulled loudly

Harsh young chicken judgement

Loud Shrimp emergency exit route

Vague pepper chicken threatens the broth

Chicken on fire complicates spice trend

Superficial clam covers the salted pond

Fresh pincer seafood ejaculation

Powerpoint Chicken

Fried intestine’s best friend lies to plumber

Breakdown in hostage negotiation squid

Fresh beef derived from Chancellery and installed in Prime Minister’s cabinet

Bon appetit!

Machine translation may be here to stay, but it surely isn’t here to replace the subtlety and intelligence that informs a high quality human translation.  How could it be?  You might as well try to start the engine and drive away in a crashed Ferrari coffee table.



For more information about Wolfestone services:

Document translation servicesLocalisation servicesTranscreation servicesMultilingual SEO servicesProofreadingVoiceover servicesInterpreting servicesMultimedia servicesLegal translation servicesOther types of translation

The professional translation services you can trust!

by Wolfestone Admin
  • Paul Chammings

    A welcome return to form, David. Yes, machine translation will have a part to play in the future of localisation and yes, it would be deeply foolish to eliminate the human factor.

    Interesting cuisine. “Fried intestine’s best friend lies to plumber” looks more like a cryptic Times Crossword clue than a meal, but I have a feeling I’ve seen “unconscionable lettuce salad fries the wheat field” on a menu in either China or Hong Kong. I seem to recall that it looked and tasted a lot like black bean corn and lettuce salad.

  • Hi David, I really love the analogy of the crushed Ferrari coffee table. It is absolute a true depiction of a situation that could have been funny if it weren’t so true!

    Here is an example: I speak German as a 3rd language, in addition to English (1st) and Afrikaans (2nd). I entered a very simple English phrase “I love buying cookies from the bakery.” and requested a translation of this sentence to German in Google Translate.

    In spite of my German being less-than-perfect-and-somewhat-broken, I immediately noticed that although six of the seven resulting words: “Ich liebe Kauf Cookies aus der Bäckerei.” can be recognised as German, the entire phrase is so horribly wrong that it will convince a German client, that you either a) have no idea what you are talking about or b) that you lack the necessary respect for him/her (as a client), to at least ensure that you express something really basic in the correct way.

    The translation should have been more along the lines of: “Ich kaufe gern Kekse in dieser Bäckerei”. I am sure your Wolfestone translators can express this way, way better than I do with my limited German abilities. 🙂

    Frankly, DIY translation is unlikely to buy trust – it is also a great way to ensure that one does not end up doing business in other countries at all 🙂

    Thanks for yet another great blog. The Wolfestone ‘Weekly’ is on my list of favourite, favourites 😀

  • Amita Sharma

    My experience of machine translation in India is that we are “catching up” with the USA and Europe which was making developments in this field for 20 years before we did. This has some advantage as we can learn from the mistakes of others perhaps. There has been a big government effort to develop proper lexical resources for these Machine Translation projects and they will become more common, but I agree with the points that David and also Debbie Bouwer is making. Namely that this is a culture issue and particularly when there is the relationship to consider between cultures.

    I have said before on these pages David that I have mixed experience of dealing with the patrician English speaker who can appear condescending to the Indian English speaker. I am concerned that machine translations which can not possibly allow for cultural issues will not only mistranslate but also do damage to relationships, and I have seen examples of this in English to Hindi and also English to English.

    What you say about the WordWright is also correct because a localised English to English copy service shows respect for the recipient of the message. Clearly this is advantageous.
    I am expecting in perhaps 2 or 5 or 10 years that a translator will be an editor of the work of a sophisticated machine, but in my lifetime I am never expecting or wanting a service which involves no human intelligence at all.

    • Well put, Amita! I agree with you on the point of not wanting a service without human intelligence behind it. Call me old fashioned, but I really am uncomfortably with the notion of machines ‘taking over’ communications between humans. *A string of disturbing Sci-Fi scenarios are crossing my mind as I type…*

  • Nicole Schiller

    Hi David! Thank you for this very interesting blog. There is no doubt that machine translation has potential uses and that human translators should not underestimate it. One undeniable advantage is that machine translation helps to avoid mistakes that human translators might commit, for example missing out on a plural. Also if you enter company-specific terminology, the system will always use it correctly.

    The fact that the weather forecast in Canada has been automatically translated from English into French and vice versa since 1977 also underlines this point. If technical writers pre-edit their texts and stick to certain rules, for example keeping sentences short and unambiguous, machine translation systems are surely able to translate these texts faster and maybe even more accurately.

    On the other hand, it is common knowledge that translation is not only about transferring words from one language to another. It is also about mediating between cultures and creating authentic texts that are not recognized as a translation. As I see it, only human translators are able to do so and capture the entire meaning and spirit of a sentence, capturing possible irony, implied messages and statements with potential for cultural misunderstandings.

    Looking forward to your next post!

  • Silke Lührmann

    Speaking as a translator I must admit that I have my weak moments of complete and utter heresy, which leads me to think I would be quite happy if I never had to translate another instructions manual, EU regulation or divorce decree in my life. Let’s leave all the dull stuff to the machines and focus on the interesting and creative jobs, on the very things that make us human – art, literature, toy names! (Yes, I do realise that this amounts to cutting my own throat, professionally speaking, since there is simply not enough interesting and creative stuff to go around – but hey, who ever said translators were entirely rational?!)
    However, the real problem with machine translations – again, speaking from a translator’s point of view, and with a translator’s economic interests at heart – is that increasingly, translators will be forced to accept “post-editing” rates for work that essentially amounts to re-translation of bad machine translations. This is already happening now, and it’s going to get a lot worse in the near future.