Google, Machine Translation, and Literature

The other week, the New York Times ran a piece on advances in Google’s translation tools, focusing on the way Google essentially crowdsources its mechanical translations by searching its mammoth database of web pages, books, etc.

Creating a translation machine has long been seen as one of the toughest challenges in artificial intelligence. For decades, computer scientists tried using a rules-based approach — teaching the computer the linguistic rules of two languages and giving it the necessary dictionaries.

But in the mid-1990s, researchers began favoring a so-called statistical approach. They found that if they fed the computer thousands or millions of passages and their human-generated translations, it could learn to make accurate guesses about how to translate new texts.

It turns out that this technique, which requires huge amounts of data and lots of computing horsepower, is right up Google’s alley. [. . .]

“This technology can make the language barrier go away,” said Franz Och, a principal scientist at Google who leads the company’s machine translation team. “It would allow anyone to communicate with anyone else.”

Statements like that fired up a number of translators, sparking at least one letter to the editor, and several snarky email exchanges. What really pissed everyone off though was this chart, which compares Google’s translation to published ones, never once mentioning the living, breathing translator’s name at all (instead referring to the “human translation”), nor acknowledging that, yes, Google can seemingly perform translation wonders when it’s searching the web for some of the most famous opening lines in the history of literature. Only a dummy would be surprised to see Google nail the opening to Gabriel Garcia Marquez’s One Hundred Years of Solitude, since Gregory Rabassa’s brilliant rendition is available virtually everywhere.

Fast forward a couple weeks, and welcome renowned translator David Bellos for the smackdown.

Bellos’s op-ed piece in yesterday’s Times is the perfect example of how to write something like this. The piece is brilliant from opening to finishing flourish. In a very balanced, smart way, he starts by describing the history of (and potential need for) machine translation, and building from there to explain the paradigm shift from thinking as language as a “code” made up of a lexicon and a grammar, to the statistical approach, which functions because people tend to say the same things over and over again in all languages.

All that’s fine and good—machine translation can help interpret when people are calling for help, when they’re making basic statements. But the implication beneath the original article (and especially that damned chart) is that machine translation can translate anything from menus to distress calls to works of high-literature. And it’s that last category which caused everyone to spit out their morning coffee. For a few reasons:

Can Google Translate ever be of any use for the creation of new literary translations into English or another language? The first thing to say is that there really is no need for it to do that: would-be translators of foreign literature are not in short supply — they are screaming for more opportunities to publish their work.

But even if the need were there, Google Translate could not do anything useful in this domain. It is not conceived or programmed to take into account the purpose, real-world context or style of any utterance. (Any system able to do that would be a truly epochal achievement, but such a miracle is not on the agenda of even the most advanced machine translation developers.)

However, to play devil’s advocate for a moment, if you were to take a decidedly jaundiced view of some genre of contemporary foreign fiction (say, French novels of adultery and inheritance), you could surmise that since such works have nothing new to say and employ only repeated formulas, then after a sufficient number of translated novels of that kind and their originals had been scanned and put up on the Web, Google Translate should be able to do a pretty good simulation of translating other regurgitations of the same ilk.

So what? That’s not what literary translation is about. For works that are truly original — and therefore worth translating — statistical machine translation hasn’t got a hope. Google Translate can provide stupendous services in many domains, but it is not set up to interpret or make readable work that is not routine — and it is unfair to ask it to try. After all, when it comes to the real challenges of literary translation, human beings have a hard time of it, too.

Well played, David.