MACHINE TRANSLATION – THE ART OF NOTHING

There is more and more talk about MT (machine translation) these days among translators. I’ve heard a lot of arguments on the pros and cons of MT, but typical of the even-handedness with which translators–who, after all, are mediators by nature–tend to discuss issues, these arguments are usually very objective.

Me, I get a singularly subjective rash and feel my blood pressure climb every time anybody talks about letting a machine translate for them, especially when the text involved is anything even vaguely “literary.” Call me old fashioned, but I consider translating a form of writing and I consider writing, at its most mundane, a craft and, at its most sublime, an art. Hence, any suggestion that I, as a translator, could use an MT program to translate a poem or a short story, is tantamount to saying that Neruda or Hemingway could have used a machine writing program to create the originals. And that is simply ludicrous.

Am I saying that there will never be a machine that can write a book? No. In fact, machine writing already exists. Perhaps the best example is the brainstorm of a fellow about whom Naom Cohen wrote an article in The New York Times a couple of years ago: namely, Philip M. Parker.
Parker is a professor of “management sciences” (whatever the hell that means), who conjured up some computer programs with which to compile data into book form. He then proceeded to “write” 200,000 books (yes, 200,000!), later having the audacity to dub himself “the most published author on the planet.”

In his NYT article, Cohen explains that what Parker has done is to develop computer algorithms that collect “publicly available information” on a subject (any subject, apparently, from medical conditions and treatments to tufted washable scatter rugs and bathmats… I kid you not), and, “aided by his 60 to 70 computers and six or seven programmers,” turns the results into books in a range of genres. They average 150 pages and are only printed when a customer buys one. In other words, other than how to program a computer to extraordinary effect, Parker would appear to be an “author” who doesn’t have to know squat about anything, nor, presumably, does he have the slightest idea whether what gets published under his name is worthwhile or even accurate–since, surely, not all of what is “publicly available” on any topic is worthy of repeating or useful as a reference. And if the computer–which, let’s face it, is mindless–is the one calling the shots and the “author” knows next to nothing about many of the subjects covered, then one can plan on having to wade through a lot of subjective claptrap and unmitigated bullcrap before actually discovering anything worthwhile reading, let alone recalling, about books “written” in this way. This, then, is an attempt at turning “culture” in general and “writing” in particular into a commodity.

Most translators would probably get this if you were to explain it to them from the standpoint of writing rather than translation. The thought that any worthwhile literature can be created by a machine would probably seem like a travesty to even the most tech-minded of translators. But when it comes to machine translation, translators are starting to fall prey to the hype. And the marketing is being so cleverly developed by the companies and wholesale translation agencies that are behind not only the software sales but also the use of such programs in creating translation memory banks designed to pay for less and less original translation, that anyone refusing to incorporate such “tools” is treated, increasingly, like a troglodyte and a social pariah in the progressively technified mainstream translation community. Translators are slowly but surely being brainwashed into believing that MT is a tool that is being created for their convenience, to alleviate their workload and to permit them to take on bigger and bigger assignments all the time and thus, presumably, make more money. The truth is, however, that the more translation becomes “commoditized” the lower the rate per word will fall. It’s a simply matter of supply and demand.

But it is also a matter of quality, since this is like trying to convince yourself that the quality and effectiveness of a twenty-dollar mass-produced wristwatch are any match for the craftsmanship, complications, calibers and excellent materials of a limited edition Swiss watch. The two simply cannot be compared. Nor can machine translation or machine book-writing be compared with the highest quality standards for the professional writing and translating craft. As with the comparison of a massed-produced watch to a fine timepiece, they are two entirely different things.

For some types of technical and legal translation I can understand why professional translators with heavy workloads would be tempted to make use of certain electronic tools to help them quickly get through highly repetitive texts, to create permanent glossaries or to improve term consistency. But, bottom line, I think, any sort of translation tool (including the myriad dictionaries we use) is only as good as the professional who is employing it.

In literary and journalistic translation, which is the majority of the work I do, I have found no substitute for straight brain to page translation. To my mind it is laughable, if not insulting, for anyone to suggest that this can be done any other way. As insulting, in fact, as saying that a machine can write “just as good a book” as a seasoned writer. So depending on and trusting software to do a job equal to or better than the translator’s own brain seems to me at least fanciful if not dangerous.

All good translation requires writing skills–the possible exception being highly technical documents that mostly consist of lists or diagrams. On the other hand, however, I once worked for three years for a nuclear technology firm, translating its training and instruction manuals for a turnkey plant project. It was there that I learned how important clear, concise, understandable writing is on the job site. That was, in fact, why they hired me, a non-techy writer, to do the job. The point being that I was capable of learning the technology, or enough so as to be able to thoroughly understand what was being said, and to translate the message into clear, concise, understandable prose. That was something they hadn’t been getting from their former technical translators and it was the reason why they kept me on until the project was finished.

When it comes to literary work, there is simply no substitute for proper research and writing skills. Last year I translated a book that, because of the complex nature of the subject matter (which covered theology, philosophy, world history, international organizations, major treaties and politics) required two to three hours of research for every hour of actual translation. Fortunately, the author was an intellectual of substance, who understood this and understood the nature of the translator’s craft. He wasn’t interested, then, in getting the translation fast, but in getting it right. What this meant was that I, as the translator, had to know almost as much about his subject as he did.

The project ended up taking me nine months. The thought that I could have done less research by using MT software to “help me out” strikes me as hilarious. And in such cases, I doubt there will ever be a substitute for applying the seat of your pants to a chair for hours on end until the job is done, and done right.

Dan Newland
Originally posted in Dan’s blog
Ilustrated by Juan Tavela