There’s Apple’s Siri, of course, and Amazon’s Alexa, and Google’s digital assistant, too. But Cortana may soon become the best virtual concierge, if only because she can actually understand you.
Microsoft said this week that it has developed new speech technology which the company claims transcribes conversational speech at a human level of quality.
Calling the achievement “historic,” Microsoft’s chief speech scientist Xuedong Huang says the company has reached “human parity” with regards to digital speech recognition.
The system’s word error rate is around 6%, roughly in line with professional transcribers, according to Microsoft. The technology will be integrated into Cortana, the company says.
I recently visited Microsoft’s headquarters in Redmond, specifically the facility for their research arm, and it was quite evident that the company takes speech recognition very seriously. Don’t be surprised to see Microsoft continue to lead this space.
That said, there is much more to effective speech recognition than simply understand which words were said. Factors like context and emotion greatly influence the meaning of words, for example, and it’s these myriad nuances of human language that artificial intelligence remains far from comprehending.