Oct 14, 2012

Who's afraid of the BDÜ?

A recent publicity stunt by the German Association of Translators and Interpreters (BDÜ) has provoked some interesting responses, most of which, I think, reveal the personal agendas and prejudices of the respondents and miss the point entirely. In a brilliant bit of political theater, financial translation wizard Ralf Lemster was pitted against The Machine loved and feared by many: Google Translate.

Well-known MT pundits like Kirti Vashee know this isn't "fair" and uttered the expected objections and protests. I imagine the details of their criticisms are correct, or largely so. But they are also largely irrelevant. They are all too aware of the Imperial Elephant in the room: public perceptions and how too many private individuals, corporations and - to their great shame - language service brokers apply machine translation in ways that are entirely inappropriate.

All the talk of controlled language, fit-for-purpose, quality-is-what-the buyer-thinks-it-is and "it's here to stay, so deal with it" and all the other noise generated by MT's sycophant choir reminds me a bit of the old carnival shell game, except that the ones taking your money there are a bit more honest in their game than ones who tell you now to "get on the MT boat or drown".

Off-hand comment? Bollocks. It was a prepared keynote speech at memoQfest 2012 and carefully calibrated to play on the fears, uncertainties and doubts of listeners and drive them down the drain of the post-editing sewage cycle. It is a far cry from the restrained, responsible approach to machine translation promoted by technologists I know who understand the extremely limited scope of MT and never oversell its performance or potential (at least not within my hearing).

The MT carnies know instinctively why the ludicrous BDÜ "case study" is brilliant. It is short of fact and science and follows a familiar lightweight format we all know from modern "news" reporting. And however much better custom, optimized MT engines are said to be, it is Google Translate more than any other which casts the biggest shadow across the translation landscape.

I remember a few years ago how Kilgray once objected on ethical principle to officially supporting a Google Translate plug-in for memoQ, because of the very real violations of confidentiality and often law which occur when translators send their clients' content to Google for processing. This promiscuity with the intellectual property and privacy rights of others is not excused by the frequency of its practice any more than the widespread practice of unsafe sex in countries devastated by  double-digit AIDS infection rates makes the risk of illicit mingling any less.

But SDL, long an ethical "innovator" in the world of translation (remember how the Trados gang brought you the Big Lie of how translators prefer to give fuzzy discounts - long before most translators had CAT tools or even knew what a fuzzy match was), apparently had no such reservations and made the tools to facilitate breaches of confidentiality more accessible to translators and wannabes, so Kilgray as well as many others gave in to popular demand and indifference to the law and released its own Google Translate plug-in officially. (Of course I support the right of SDL, Kilgray, Atril and any other company to make such tools freely available. A Google Translate plug-in no more violates intellectual property rights than guns kill people. We all know and accept that it is humans and their weak nature which are at fault, and we can do nothing about that but let things take their course, right?)

Reactions from translators who understand German were fascinating. Some pointed out that the New York Times had already "gone there" (not really - the translations of literature snippets are even less relevant I think... I want to see real-life risks shown with electrical repair instructions and instructions for surgery or the use of pharmaceuticals), others criticized the BDÜ for not showing side-by-side comparisons of the machine's erroneous spew and Ralf's correct text. But I think the Devil's distraction is in the details.

The BDÜ is generally rather hidebound and out of touch with many aspects of modern translation practice; the pages of nonsense in the online registration for their recent conference in Berlin tried my patience to the point where I decided to stay home and make jam for the second time. But this time they got it all right I think.

A good guerrilla knows that the battle is about hearts and minds, and since most of the public is a bit short in the latter capacity, it's best to go straight to the heart of the matter with an entertaining show. A week after watching the "report", few will remember the details. But they will remember the emotions evoked and the air of authority projected. MT will lose every battle of fact on the fields where its carnies pitch their tents and the public crowds gather, but the shysters have a very sure advantage and exploit it at every opportunity: the naive confidence of the scientifically illiterate that "progress" will continue and things will always get better, MT included. I hear this all the time from colleagues. The less they understand the technology they use, the more they are willing to be used by it, and the greater their confidence that it will inevitably infiltrate their necessary professional activities.

A similar confidence once existed for the impending discovery of the Philosopher's Stone to turn lead to gold, occupying even the greatest minds like Sir Isaac Newton. Skeptics like me were most likely laughingstocks to go against such wisdom.

The way to win this "war" is not with facts to be forgotten in a day. The path to the victory of good sense and the human spirit lies in a bit of theater like the BDÜ has offered and perhaps some good jokes like the late Miguel Llorens so generously shared at the expense of dishonest MT promotion.

That is something which the carnies have good reason to fear.


  1. One thing I think should always be pointed out to the proponents of MT+postediting as a recipe for speedy and "good enough" translation: who is going to do the postediting?

    Because, since good translators won't touch it, what remains is MT+postediting done by inexeprienced and/or unskilled translators. The most likely result, in most cases, at least, is *adding* human errors to the machine's ones.

  2. I don't get it from both sides. I don't get the MT people who think they can sell MT in a world with free Google, and I don't get the trad crowd who refuse to use MT (or even CAT) when it could speed up an ease their production. So when was the last time you translated the word "Tuesday"? And when was the last time you translated the N-gram "with regard to the following"? Tools are a means to an end product. Bad carpenters blame tools. Good carpenters use tools appropriately. Non-carpenters don't even know what a table is. No worries - there'll be no dinner to eat off it....

    1. "I don't get the trad crowd who refuse to use MT (or even CAT) when it could speed up an ease their production."

      First of all, I think you are conflating two different things: I know there are translators who still refuse to use CAT tools, but outside translation for the publishing industry, and certain other niche fields, they are a dwindling minority. Most professional translators now do use CAT tools, both because they help us increase our productivity, and because they help in other ways as well (maintaining consistency in our translations, for example).

      There are many more of us, on the other hand, who do not use MT... because we find that unlike CAT tools, MT does not, in fact, much increase our productivity... and also (and especially) because while CAT tools, when used properly, tend to increase translation quality, the use of MT is sure to sharply decrease quality.

  3. Great post, Kevin. Couldn't agree more with you, Riccardo.

    @Anonymous: what type of MT (>>>post-editing) are we talking about?

    1) MT for one's own work (post-editing of one's work)
    Although I have not done it so far, I might consider using an open source software in MY computer or buying a Systran license (offline, of course) for its obvious advantages. In any of the two cases, the benefit will be for me as everything will "happen" in my own computer. I don't think many translators in the world would oppose to this use of MT. For personal reasons, you may not be willing to use MT (in YOUR computer for YOUR business), but it is clear that this use of MT is not negative for the profession. (Same with TMs, once I realized the benefits I could get from them, I bought my Trados license. And the benefits of my investment are... mine.)

    2) Post-editing for agencies
    (Huge) however, when it comes to post-editing for agencies, the situation is totally different. I will leave my profession before post-editing for an agency (no need to explain what the MT+TM combo represents for the profession).

    And a huge 'no' to online MT, of course, as it is illegal to disclose our clients' documents (without clients authorization).

    Nobody is against technology, it´s just that we got tired of others profiting from technology while pressure on us (the trad crowd, as you call us) has been increasingly hurting our business.

  4. Hi Kevin (and others)

    can we think a bit more about the concept of "violations of confidentiality"?

    imagine that the renowned Doctor X just invented the cronovisor and it works

    the ACME company acquires the patent and decides to sell it in the USA, firstly

    after a great success, ACME decides to sell it in the whole world, so hires a Multi Language Vendor to translate the Instructions For Use in a lot of languages, Italian included

    if, now, an Italian translator puts the English text into GT he/she doesn't break any confidentiality IMHO, as those IFUs come from a widely sold and used tool !

    and, frankly, in 12 yrs as a translator, I never seen (even if I really wanted to see) an IFU associated to a brand new stuff/item

  5. Yes, Claudio, let's think.

    Perhaps the translation is of the initial, internal concept disclosure - from a subsidiary's research department in Italy and intended for understanding by decision-makers at a parent company. Perhaps the translation is of a proposed legal settlement for an injury case, the terms of the settlement being strictly confidential and containing personal details about *your* daughter. I could think of many other situations where the use of Google Translate might indeed violate confidentiality. The fact that you don't feel you've seen that translating instructions for commercial devices doesn't change that.

    And if you feel that GT improves the quality of your text, for example the clarity of your expression, make use of the tool at every opportunity, by all means, for your non-confidential texts ;-)

  6. Dear Claudio,
    Unless your client authorizes it, using GT is not permitted. I commented further here:

  7. An issue with the real practical world and how many of us are taking these features seriously unless and until, it is felt apart through the great search engine. No doubt, Germany is the Retail Giant and every corner of the world sales is directly and indirectly associated with the communication factors.

    As per the Confidential and Violation tales, tools and apps need improvisation at regular minutes to ensure an honest path leading to a better world in every sector of operations. All visitors would agree- Greater the Confidence certain the Infiltration during Professional activities.
    German English Translator


Notice to spammers: your locations are being traced and fed to the recreational target list for my new line of chemical weapon drones :-)