Google’s Translation Phone Not Imminent

Contrary to media reports, Google is not about to launch an automatic translation phone, but the company is working hard on the two underlying technologies that could make such a device possible, a spokesman told eWEEK Europe.

“We think speech-to-speech translation should be possible and work reasonably well in a few years’ time,” said Franz Och, Google’s head of translation services, in an interview which the Sunday Times titled “Google leaps language barrier with translator phone”, although Och made no such announcement.

Other sites leapt on the story, including the Daily Mail which advised readers “don’t bother to learn foreign languages“.

In fact, Google is merely working hard, alongside other organisations, on the two underlying problems facing anyone wanting to build a translator phone: machine translation and voice recognition. “We think it is possible, in a few years, and we are working on the two things that would make it possible,” the Google spokesman explained to us.

“Clearly, for it to work smoothly, you need a combination of high-accuracy machine translation and high-accuracy voice recognition, and that’s what we’re working on,” Och (left) told the Times. “If you look at the progress in machine translation and corresponding advances in voice recognition, there has been huge progress recently.”

Speech transcription, in particular,w has proved a tough nut to crack. While PC software that is trained to a specific user has become useful, services that will work with any voice have proved elusive.

In July last year, speech transcription company Spinvox faced a scandal when it was alleged that its automatic transcription service actually relied on human operators. The company was later bought out by the leader in that field, Nuance.

Google’s work apparently takes this difficulty into account, and is concentrating on translating the phone owner’s voice, rather than that of another user. “Everyone has a different voice, accent and pitch,” Och told the Times. “But recognition should be effective with mobile phones because by nature they are personal to you. The phone should get a feel for your voice from past voice search queries, for example.”

Excitement over a possible voice-translating Google phone is understandable, given other recent activity by the search giant. Google is offering a business version of its Google Voice Internet telephony service.

Meanwhile, Google’s Nexus One phone, which is causing excitement worldwide, includes voice-recognition features. These features do not work very well, even with the user’s own voice, as our review commented, though the phone’s noise-cancellation features might provide a good enough sound signal to start working on speech transcription.

David Crystal, a professor of linguistics at Bangor University who is quoted as a sceptic in the Times article, is critical of the piece: “I didn’t actually say what he says I said,” he said in a comment to the article. “I said that there are two problems preventing successful automatic voice recognition, not one: regional accent diversity and speed of speech.”

Professor Crystal also poured cold water on Mail readers’ hopes: “Nor did I say that the need to learn foreign languages is removed. I believe the need to learn languages will be stronger than ever, if and when such software becomes available, because there are all kinds of benefits which come from learning a foreign language. The point is being repeatedly affirmed in current research, and it’s a pity the opportunity to reaffirm it wasn’t taken up here.”

Microsoft – whose Windows Mobile 7 is struggling with delays – has also talked about speech-recognition as a better way to interact with a PC or smartphone.