Machine Translation: je t’aime… moi non plus *

Had blogs existed with the advent of the ATM machine, bank tellers would no doubt have spilled many bytes of virtual ink in an effort to disparage them. And then, little by little, a few may have come around to extoll their virtues, claiming that the new machine would free up precious time from the mindless task of counting bills and let them get on with the more human aspects of their job.
I won’t take a position on whether machine translation (MT) poses a grave threat, or represents a benefit, to the translator. I can’t predict exactly when the various tools created by computational linguists will get it right. Nor can I say with assurance how many tellers successfully transitioned to a more rewarding gig at the bank.
I read two interesting posts recently, both pro-MT, yet seemingly on two ends of the spectrum. Martin Cross explores Google Translate and although he has misgivings about this specific program, he begins on a very upbeat note about machine translation in general. Far from the ATM scenario, his view of the threat is akin to how “a litigation attorney might see a website that will write your Last Will and Testament for $9.95.” Martin goes on to assert that MT will produce more work for translators because it will unearth foreign documents nobody would have looked at to start with, and some of these documents will need a human touch. I like his fearlessness. And his logic. I had never thought of it that way. If true, our golden age is still ahead.
A.M. Sall, on his terrific blog, sees the future differently. In his post, “Machine Translation - A New Convert :-)”, he claims that the good translator will go on to post-edit (PE) translations produced by machine. This MT PE model is already in the marketplace, according to Sall, and being pioneered by Lorena Guerra and her company Euromix. It looks from the website that they may not have begun the MT PE system in ernest quite yet; I find only the standard human version offered.
Sall tells that according to Jeff Allen, Director of Business Development at Translations.com, a machine translation and a post-edit can be achieved in one-third the time. But this requires, in Sall’s words: “an excellent translator, which means perfect linguistic competency, supreme cultural fluency and sound subject matter expertise.”
I’ve always wanted to think I was one of these Super Editors, having edited many pages in my career, but alas, I’m not confident I could live up to those conditions, and I see the cost payoff Sall alludes to somewhat far off in practical terms. I have placed many jobs with translators and editors over the years and find that, in general, (1) translators prefer not to edit, and (2) when they do, they care very much who did the translation. But who knows, maybe when machines improve, editors will have warmed up to the idea.
I reasoned in Martin’s blog that the biggest difficulty to overcome with machines was a lack of logic. And thus, for a client, a lack of reliability. We know that human translators make blunders, too. Both machines AND humans can be unreliable translators, but they are unreliable in different ways. A human slowly assimilates the narrative and context of a document, and this larger picture informs the small puzzle pieces along the way, and vice versa. A machine that translates, though it may be “learning” depending on the tool we’re talking about, does not think its way through the text. Thus, based on knowledge of the big picture, a human translator will be able to answer when we ask, “I don’t get it, what does this sentence mean”? I believe the machine will give us back the same sentence.
Now, the fact that computers appear to be getting closer in their approximation to thought and human language is another story, and it may soon prove close enough to be practicable. But I recently listened with some interest to an interview of a robot scientist who claimed that for all the progress in robotics, they still haven’t been able to develop a model that’s as smart as a lobotomized cockroach. Somehow I felt reassured.
* [tr: I love you... me neither]: title of a song by Serge Gainsbourg.










2 responses so far ↓
1 The GITS Blog » So it’s not just me // May 22, 2008 at 2:10 am
[...] yndigo blog confirms that I’m not alone in not liking to edit others’ translations: I have placed many jobs with translators and editors [...]
2 Machine translation and gisting | yndigo // Jun 5, 2008 at 3:06 pm
[...] The merits of various MT programs or MT in general have been discussed for decades. Today, translators seem to fall into three camps: the first says that machine translation is decades away from being usable; others claim it will soon put us all out of work; still a third group believes it’s ushering in a new golden age that will uncover mountains of previously undiscovered text for human translation and generate even more work fine-tuning software and “post-editing.” I explored some of these ideas in an earlier post. [...]
Leave a Comment