Google Working On Real-Time Language Translation Phone - HotHardware
Google Working On Real-Time Language Translation Phone

Google Working On Real-Time Language Translation Phone

Google has a history of removing language barriers, but the company hasn't yet made its web-based translation service especially easy to access or use from a mobile device. We knew that Toshiba was actually working on a translation phone, but now the same company responsible for Android and the Nexus One is apparently doing the same thing. Anyone want to guess how this will end?

According to a new report Google is hoping to have "a basic" translation system for mobiles in place within a couple of years, and it'll be far more advanced than some app where you still have to punch in a phrase or URL in order to get something translated. Basically the idea is to combine Google's voice recognition technology with its web-based translation service, so that the phone itself could recognize the language being spoken in real time and then "translating it into a synthetic equivalent in a foreign language."



If this works, you could theoretically speak to someone on the phone who doesn't speak your language. And you two could communicate in real-time. Something tells us the iPhone App Store won't be getting this right after Android, but who knows.
0
+ -

Now this is a cool thing to see happening. Google has to have stolen someone's idea gnome. I have thinking of that statement noticed some in the industry who have definite lost there's. Of course some companies switch to another part of the industry, some disappear, but Google always not only comes up with new ideas, they always seem to be good ones as well.

0
+ -

>Something tells us the iPhone App Store won't be getting this right after Android, but who knows.

Reallly? I seem to recall a HotHardware news story that indicates the iPhone already has it:

Jibbigo: A Real-Time English/Spanish Speech Translator For iPhone

It was written by a fellow named Shawn Oliver. Perhaps you know him.  .)

Speech input has a variety of problems. One, of course, is discernment of the human voice-- our "processors" (i.e. temporal lobes, specifically Broca's and Wernicke's areas) are incredibly flexible, and yet we still fail when in high-noise situations-- even cocktail parties, where the mere mention of our name in another part of the room can draw our attention away from a conversation right in front of us. (Fascinating effect, really-- look up "Cocktail party effect.") A computer's audio input is almost always monaural, so it's listening without the advantage of stereo, and the improved focus that gives.

(Hey, how about using face-recognition technology to read lips? I hereby call this idea. Anyone who develops it needs to pay me lots of money.)

Even with great listening abiity, computers haven't gained the ability to work around human speech patterns. Think of a Tom Stoppard play, compared to an Oscar Wilde play: Wilde wrote in complete (and well-thought-out) sentences, and any actor worth his greasepaint will say them in a clear and unhesitating voice. It's great to watch. It isn't the way real people speak in public, though. Stoppard writes in a more familiar fashion, inserting the "ums," "ahs," and other verbal hesitations; his characters stop a word in the middle to start a parenthetical statement, then pick up where they left off. Humans can understand that perfectly. Programs can't (at least for the moment).

But let's assume that you can get a decent signal, you've trained yourself to speak briefly and to the point (perhaps you took up Quakerism), and enjoy great recognition, more than the 85% often cited. A good translator won't just give you one version but several, based on the likelihood that the context of the words was what it thought. I remember translating Japanese, and seeing lots of references to a "temple face." Later I realized that "Shutaro Mendou" was a name. (The writer, Rumiko Takahashi, enjoys punny names, which are anathema to a translator program.) But a voice output implies just one result being chosen-- and if it's the wrong shade of meaning, well, let's just say that it wouldn't impress a French girl.

So, it's not up to snuff for serious conversation. But it should be able to help you translate commonly used handy phrases like "Where is the bathroom," "Please take me to the airport," "My hovercraft is full of eels," and "How much is this rug?" Congratulations! You've just spent $300+ to do the same thing that a $2 laminated wallet card will do for you.

0
+ -

Thanks for the info ClemSnide. I thought this tech seemed familiar, though Jibbigo is limited to Spanish and English.

Does this remind anyone of the Universal translator from Star Trek?

0
+ -

Mmm, more like the voice capability of the rides in Firesign Theater's I Think We're All Bozos On This Bus.

"I'm going to say something and (chuckle) you're not going to be able to understand it. Why does the porridge bird lay its egg in the air?"

"White does the poor-rich burr lazy leg in the ire?"

Login or Register to Comment
Post a Comment
Username:   Password: