MOUNTAIN VIEW, Calif.--One of the hits at this year's TechFest in Redmond was the so-called translating telephone.
The effort combines speech recognition, machine translation, and text-to-speech technology to let two people who don't share a common language nonetheless carry on a phone conversation. As part of a regional TechFair at Microsoft's Silicon Valley offices on Thursday, I got a chance to try out the technology, conversing with Frank Seide, one of the researchers behind the project.
I asked a few questions in English, while Seide answered in German. As one can see from the video below, it's far from perfect, but even being able to get the gist of what someone is saying without sharing a common language is pretty cool.
Speech recognition and machine translation alone are prone to errors, so certainly combining the two is likely to lead to some pretty comical results, but it certainly shows the promise such technology holds. Even today, it would probably do in a pinch.
Not all the errors were the computer's fault either. In a couple of cases, the translation or speech recognition engines didn't get things quite right. However, another time, it was a case of human miscommunication. I asked Seide how long Microsoft had been working on the project and, instead, he explained how the project worked, which the prototype then translated.
For more cool stuff from TechFair, check out this story on how Microsoft researchers are creating some pretty decent still images from some rather wobbly video.