I’ve never Skyped with someone speaking a foreign language — unless you count the students I hold Skype office hours with. But it makes sense — I’d rather Skype than call someone at the front desk of a hostel to make a reservation, or to set up a scuba diving expedition in Thailand, largely because it’s so much cheaper. Anticipating that Skype will continue to replace phone interactions among international users, the company will soon be testing “Skype Translator,” which offers real-time translation.
The test version follows a major breakthrough for Microsoft in speech recognition. Figuring out a way for computer systems to understand speech has been a goal of computer scientists for decades. The first approaches involved pattern recognition and matching, linking waveforms to words, but given the difference in people’s voices, including accents, inflections, and speech patterns, the technique didn’t produce consistent results. In the 1970s, researchers began amassing databases of speech models that help improve speech systems to the point where they are now — that we can call in to companies and “talk” to an automated system that recognizes what we’re saying (though, sadly, doesn’t respond whatsoever when we swear at them out of frustration). But even that’s not good enough for an internet phone service.
Two years ago, the University of Toronto and Microsoft teamed up to design something called Deep Neural Networks, which does more than recognize speech patterns, it actually recognizes human brain patterns. The technique has resulted in a 30% drop in the spoken-word error rate of automated systems, such that they can correctly recognize an average of seven out of eight words. With additional data and training, that rate will improve.
Those of us who are looking for a real-time translator have a few options, including Google Translate, a free app that translates signs and other text picked up on the user’s camera, but that’s a far cry from a conversation. Skype’s 300 million monthly users will have access to the service, though it’s not known yet whether it will involve an additional fee. The move with make Skype more competitive against Google Hangout and Apple’s Facetime, among others. Skype also has 3D video calls in the works.
Skype’s corporate vice president said, “It is early days for this technology, but the Star Trek vision for a Universal Translator isn’t a galaxy away, and its potential is every bit as exciting as those Star Trek examples.” Maybe Skype’s new service will help us figure out what William Shatner — and my students — are talking about these days.