Can you imagine traveling to China and speaking to people in fluent Mandarin with no prior knowledge of the language? According to software giant Microsoft, this could soon be a reality.
The Microsoft research team is currently developing and refining speech translation software that is capable of translating speech instantly. The technology imitates the intonation and cadence of the speaker, delivering more real and natural-sounding translations.
In a recent video presentation, Microsoft’s Chief Research Officer, Rick Rashid, demonstrated how their translation technology converts spoken English into Mandarin – in real time and in the speaker’s own voice. Watch the demo here.
Although today there are a number of translation technologies that deal with human speech recognition, Microsoft wants to go a step further and perfect past breakthroughs.
Working with scientists from the University of Toronto, Microsoft has been able to slash translation errors from 20-25% down to 15% thanks to a technique called Deep Neural Networks. With this technique, which is modeled on how the human brain works, the researchers were “able to train more discriminative and better speech recognizers than previous methods.”
While the technology is still not perfect, Rashid calls the improvement a “dramatic change” and believes that “in a few years we will have systems that can completely break down language barriers…we may not have to wait until the 22nd century for a usable equivalent of Star Trek’s universal translator.”