A Start-Up Bets on Human Translators Over Machines

By SOMINI SENGUPTA

There may be one thing that humans can do better than machines: understand the meaning of words.

Unlike a computer, we can easily distinguish between a river bank and a savings bank. And our translations, even when they involve a language we barely comprehend, turn out to be better than what the best computer program can do. Satire? Algorithms have great trouble with that. Irony? Forget about it.

That edge in decoding what things mean is what a computer scientist turned entrepreneur, Luis von Ahn, is betting on. His start-up, Duolingo, which opened to the public on Tuesday, proposes to put language learners to work to translate text on the Web, all for free.

For the language learners, there are free lessons, translating one sentence at a time, from simple to more difficult. For enterprises that want translations for their online content, it’s free labor; a professional translator can be costly. Google Translate, by contrast, relies on machines – and can sometimes produce bewildering passages.

For now, Duolingo, which has been available by invitation only for the last five months, is limited to English, Spanish, French and German. People and companies can submit their content to Duolingo for translation, a service that the company may begin to charge for, or Duolingo could harness whatever is not under copyright or licensed under a liberal Creative Commons license.

Photo by Justin MerrimanLuis von Ahn

“You’re learning a language and at the same time, helping to translate the Web,” Mr. von Ahn said. “You’re learning by doing.”

Crowdsourcing is at the heart of Mr. von Ahn’s ambitions. His last enterprise, reCaptcha, makes use of the squiggly, wavy letters and numbers that Web users transcribe every day on Web sites to ensure that they are not robots trying to break in. Mr. von Ahn gathered those squiggles from digitized images of old manuscripts, books and newspapers – including The New York Times. Every time they transcribed the wavy words, Web users helped transcribe fading texts that were hard for a machine to read – for free. Google bought his start-up in 2009.

Mr. von Ahn came up with the crowdsourced translation idea when he noticed how friends and relatives in his native Guatemala who do not speak English have far more limited content available to them online.

The Web is just not as good in Spanish, he said. “It’s got much less information. I see people struggling with that a lot. They don’t get the information we take for granted.”

The New York Times has been experimenting with Duolingo as a potential means to translate its digital content to other languages, said Marc Frons, the company’s chief information officer, but has made no commitments to using the service.

Mr. von Ahn is thinking of taking on Wikipedia as his first project. He said, “eventually we intend to charge content providers either for faster or more accurate translations.”

For Duolingo to work, it needs a vast crowd of learners. The more proficient they become, the greater the chances of accurate translations. But even beginners, Mr. von Ahn says, get things right surprisingly often. In his program, a large piece of text is broken into easy and difficult pieces – by a computer, of course – then divvied up to students at varying levels and put back together, again by a machine.

Duolingo has raised $3.3 million. The actor Ashton Kutcher is among the backers, along with Union Square Ventures and the business advice author Tim Ferris.

NYT