Machine translation could save minority languages

We’ve seen Word Lens, which translates signs automatically. Now this:

Google Translate App works as you speak

In an attempt to break down language barriers the world over, Google have developed an App which allows you to translate your words into another language as you speak.

Users simply speak into the device and the Google Translate app translates your speech and then reads the translation out loud, all in real time.

The person you are conversing with can then respond in their own language and their translated words will be spoken back to you.

But language teachers and linguists can rest easy that they’re not about to be put out of a job just yet: currently the app only supports English and Spanish.

I’d love to play with a copy. At the moment, I suspect it’d be pretty rudimentary and error-prone, but there would be updates. The task of machine translation is as yet unsolved — or should I say the set of problems that converge on MT — but we keep seeing innovations that get us closer and closer to that goal, inch by inch.

If we ever do realise the goal of instantaneous, unconstrained automatic translation, communication would of course be the most obvious beneficiary, but the other would be minority languages. It could potentially save them.

I see it as similar to the OS wars of the 90’s, which, like language, was a conflict over standards. Computer operating systems, like languages, require a population of users who can exchange information (in this case, files) with each other. But cross-platform file compatibility issues made this difficult. Operating systems also run applications that won’t run on other systems, so there’s a disincentive to adopt an OS that doesn’t have the software you need. At the time, the Mac was on the bad end of that struggle — there were fewer installed users, and some programs weren’t available for the Mac. I remember feeling very concerned that the Mac OS would die out.

Then the Mac adopted standards that were in common use anyway — text was text no matter what computer you were on, jpegs were jpegs, and Word files didn’t need to be converted. (Perhaps Mac users should be grateful for Word after all.) You couldn’t run the exact same programs, but every computer became able to do mostly the same things: Java, Perl, Flash, and so on. And if you got really desperate, there were Windows emulators. So the cost of settling on a minority OS went way down.

What automatic machine translation does is lower the cost of maintaining a minority language. Languages like English or Mandarin have an irresistible attraction for speakers of other languages because they have a huge install base. They represent economic and social opportunity. If translation between them is easy, then using the other language isn’t an irretrievable commitment.

You could argue that the ease of translation would doom minority languages because the translation might only flow one way: toward the big language. That’s not what happened in the OS wars. People liked their Macs, and the ease of conversion helped them hang on to them. People like their languages, too. They’re important markers of their identity. But not if the cost is too high. MT would bring the cost down.