COLUMNISTS

Mark H. Teeter

Over a half-century has passed since IBM researchers proudly proclaimed that "Russian was translated into English by an electronic 'brain' today for the first time."

In the interim, technology has advanced by leaps and bounds - man has walked on the moon and machines have beaten him at chess - and yet Russians studying English in our tech-savvy 21st century find that simple homework sentences can still come out gobbledygook when fed into a translating machine. Life isn't fair, kids.  

And companies lie. Or get optimistically confused, let's say. When the folks at Big Blue told the world in 1954 that their functional Russian-to-English "brain" had put routine machine translation on the five-year horizon, they may well have believed it. The problem, of course, was that they didn't understand the problem.

You could argue, and some do, that machine translation actually has arrived. Various systems - including Yahoo's Babel Fish, the Google Translate project and a universal translator developed by the U.S.-based Defense Advanced Research Projects Agency (DARPA) - can produce written or spoken results (in various languages) with an accuracy rated at 70 percent to 80 percent.

Sound pretty good? Well, if the 20 percent or 30 percent that's garbled concerns, say, the intricacies of a multi-stage corporate divestment or a complex surgical procedure, you may not be that excited that the machine "got most of it right" - because you may be broke or dead. Yes, there are even worse outcomes than botching your English homework.

Intricacy and complexity are the key elements here. And the two can be supplemented by perhaps five more - subtlety, discernment, nuance, discretion, and grace - to make up a formidable Gang of Seven, a pleiad of translation qualities and skills that machine work lacks, entirely or in some degree, keeping human translators like me both indispensable and solvent. OK, relatively solvent.

The absence of these qualities in a machine job is all too easy to illustrate. When you feed a common, everyday Russian phrase into Babel Fish, for example - let's say "Пошёл ты вон отсюда!"/"Poshyol ty von otsiuda!", the equivalent of "Get out of here, you!" or the colloquialism "Beat it!" - what comes back is, alas, "You went there hence!", a translation that scores 100 percent for word rendition and 0 percent at communicating the desired message. The undesired object is likely still standing before you (and possibly mumbling "Hence?").

Sure, the DARPA translator, Google and other Babel Fish applications normally show a much-closer-to-human facility than this random example suggests. But any way you parade their virtues, machines still can't threaten their current accuracy threshold very seriously: for all their lightning speed, effortless recall and unflagging attention, they still have us for a teacher.

The human brain enjoys a limited understanding of its own capabilities, as the optimists at IBM found out. Since we still can't envision all the pitfalls, actual and potential, that the mysterious translation process comprises, our ability to teach a lesser entity such as a computer how to navigate its straits remains imperfect - and always will. Imperfection, after all, is how we define ourselves: to err is human.

This fallibility extends, of course, beyond the flawed construction of translating machines. We also make errors of purely human judgment, the least forgivable of which, in the translator's realm, may be a sin of omission: Not Asking an Available Native Speaker. The city of Moscow, for example, has pointedly Not Asked me for help on several occasions.

The Metro maps in subway cars here were once labeled "Map of the Moscow Underground" - leaving millions of non-Londoners to marvel at a Russian government that not only allowed clandestine opposition, but provided directions for locating it!

Though this gaffe was soon righted - with new maps to the "Moscow Rapid Transit System" - the transportation people above ground were next to Not Ask, deciding that переулок/pereulok should be rendered on bilingual street signs not by the obvious choice, "Lane," but by the lumpy and un-English "side-street" - yes, with a lower-case s and a dash for bad measure. Do Moscow officials really think the Beatles sang fondly of "Penny side-street"?

All Mayor Yury Luzhkov has to do to head off the next clunker is ask me - or my students (after they've finished their translation homework). In the meantime, recapping today's learning points:

1. Don't hold your breath waiting for a conspicuously reliable machine translator - you'll burst. Or 20 percent to 30 percent of you will. Ouch.

2. Take a native English speaker to lunch. You'll be surprised at the expert assistance a cold beer and some chili-cheese fries can produce!

Next time: Found in translation:

Yes, there is good news!

ExtremeEnglish@moscownews.ru 

Mark H. Teeter is an American English teacher and translator based in Moscow.