There are few barriers left to break through when it comes to digital communication, but spoken language remains one of the last hurdles for both technology and innovation to surpass. We have the means to talk to anyone, anywhere — as long as all parties involved speak the same language. An obvious impasse arises when two different languages attempt to cross paths. According to Microsoft, there are nearly 2 billion minutes of conversations logged each day on Skype — between personal computers, tablets, smartphones, and television. In each of these conversations, the participants share the commonality of language. How many more conversations would take place if Skype’s 300 million users had the ability to converse with one another regardless of their native tongue?
First to Market Advantage Goes to Microsoft
Although Google and Apple have similar real-time translation operations in play, Microsoft and Skype have beaten them to the punch. Skype Translate is the world’s first real-time, universal, language translation application; the release is projected for later in 2014 in conjunction with Windows 8. The technology was recently demonstrated by Gurdeep Pall (VP of the translation application) at the inaugural Re/Code Conference and has been a long time coming.
For the past two decades, Microsoft has had a special team researching machine translation, speech recognition, and speech synthesis — all of which both Skype and Microsoft believe are key components to understanding and advancing technologies in the more personal computing era that we live in now. With the concepts of machine learning in mind, Microsoft and Skype have created a novel model, which is where the success of this translator lies.
“Now it’s not just about daisy chaining these three technologies [machine translation, speech recognition, and speech synthesis] and bringing it together,” Microsoft CEO Satya Nadella explains. “In fact it’s about this deep neural net that you build that synthesizes a model that is able to do speech recognition in pretty magical ways.” Watch the videoof the Re/Code demonstration to see the enthusiasm Microsoft has for this new app.
Machine Learning On a Whole New Level
This new neuro-model is the reason why many might compare the way Skype Translate functions to the way a human brain functions. The program learns. Nadella further explains that if you teach Skype Translate English, it will learn English. If you then teach it Mandarin, it will learn Mandarin—but in the process of learning Mandarin, it gets even better at English. Teach it a third language—Italian, Urdu or French, for example—and it becomes apt at the third language, but absolutely excels in both English and Mandarin.
Apart from Skype Translate being a quick learner, the program nails two crucial elements of language translation — grammar and semantics. You can watch the demonstration of English-speaking Pall Skype chat with his German-speaking colleague Diana Hendricks here. Keep in mind Pall does not know German and Hendricks does not know English. Fortunately, Skype Translate is fluent in both.