I’ve been a professional translator for over 10 years, and I have to admit that online dictionaries have saved me more than a few times. But as dictionaries and, more importantly, translating programs become ever more sophisticated, does that mean I will become obsolete? I sure wish that were the case, but I’m afraid I’m here to stay.
I thoroughly enjoy translating poetry. But some of the voluminous financial documents I’ve done over the years contained sinewy sentences running over a page in length and more subordinate clauses in a single paragraph than an entire Faulkner novel — enough to engender midnight bouts of despair. Next come the verbose academic diatribes of retired professors that can provoke habit-inducing behaviors. And of course, all the mundane, mind-numbing assignments that probably make up 80% of translation work: news and announcements, brochures and manuals, business letters and web pages. Luckily, translation programs like those provided by Google and Babelfish can handle the latter to a reasonable degree; they at least convey the gist of a text.
Such programs, while convenient in a casual setting, still have a long way to go before they become truly viable. The fact that their translations need to be heavily edited isn’t the problem. It’s all the meaning that is lost, especially between Japanese and English. Linguistic science tells us these languages are worlds apart. For a native-English speaker, Japanese is a level-five language, meaning the hardest to master (a distinction it shares with Arabic). Creating a program that can negotiate such vast differences seems like a quixotic dream perhaps exceeded in difficulty only by the pursuit of robust artificial intelligence.
One approach to the problem would be a program that translates by consensus or “open-source,” so to speak. Wikipedia provides the most convenient analogy. Its entries are by no means absolute, but rather refined through increasing participation in the project. A program could always map out the structure of a given sentence (remember diagramming sentences?) and use those discrete values to create a correspondence in the target language.
But perhaps it would be best to somehow begin compiling publicly available translations of identical or similar structures and finding a kind of average or typical rendering of those structures. Like Wikipedia, people could submit suggestions. Already, the website www.alc.co.jp searches the internet for all known translations of a given word. Not all the translations it uncovers are correct, per se, but it is very useful and you can generally determine an approximation through comparison.
The first problem is that even though generally accepted grammar structures are finite, possible combinations of words and their meanings within those structures approach infinity. Mapping structures, compiling known translations, refining the system — it all seems so Sisyphean, even for a computer or enormous open-source project. Obviously, scientists relish the challenge, but my skepticism remains. How do you translate "muzukashii?" Its basic meaning is “difficult,” but it often means “impossible.” Think of the catastrophic consequences that could result from that mistranslation. And that’s just an ordinary adjective.
This problem — meaning in context — became clear to me years ago when I was doing some unusual copy-editing at a translation company. A Japanese woman was writing desperate letters to an English-speaking lover who had left her. She was having her letters translated into English and his letters translated back into Japanese. For a while, I only got to see translations of the outgoing letters, and it was apparent something was wrong. Cross-cultural, cross-lingual relationships can have more than your usual disconnect, but this was uncanny.
I asked to see all the original letters. The translations were relatively faithful on the surface, but something was missing. What were they really saying to each other? Implicit meanings were being lost. More empathy (and audacity) was required of the translators. Greater interpretation had to come into play. (A key distinction is helpful here: “interpretation” pertains to extracting meanings from a text, while “translation” involves rendering those “interpretations” into another language.) All translation entails some level of interpretation, though it seems the most important ones require a lot. When interpretation is already so difficult for humans, I’m not sure how a computer or program is going to cope.
Our every interaction with someone demands interpretation, and that is mediated in nebulous ways by what we know of the person and the context. Even with an anonymous text where those bearings are stripped away, we need to interpret the tone to determine meaning. Then there’s that most modern of afflictions: irony. Throw in humor, understatement, hyperbole, satire, oxymoron and any of the dozens of figures of speech, and you have “no small task” in translating. Language can be a nightmare as much as a miracle.
People use translation programs for their own convenience but also at their own risk. Will the risk decrease? Of course. But the nightmare won’t go away. The chances that scientists will come up with something we can trust over your seasoned translator are about that same as that couple getting back together: rather "muzukashii," I would say. Ry Beville maintains a site of poetry translations at www.nakaharachuya.com. This commentary originally appeared in Metropolis magazine (www.metropolis.co.jp).© Japan Today