When you speak to your phone — whether it’s Apple’s Siri, Amazon’s Alexa or the next one in line — you want to be heard, right? Well, Microsoft says it’s next in line for proving that it has got voice recognition down pat, better than ever before.
A team of Microsoft researchers says they’ve gotten the error rate of speech recognition down to the lowest ever: 6.3 percent. That’s down from IBM Watson’s 6.9 percent error rate.
While Google announced last week that it was closer to a breakthrough with DeepMind, this news brings Microsoft a bit closer to getting computers to comprehend to the level that a consumer needs, as well as helping with Conversation-as-a-Service through tech, such as Cortana, Skype Translator and other language-related services.
The voice assistant and the whole concept around voice as a computing interface is taking hold faster than ever before. And the competition is even fiercer. Who wants Siri returning information on the country of Mexico when you’re just looking for the address of your local Mexican restaurant?
Functionality and language understanding through artificial intelligence are improving. That means better accuracy and convenience, with Amazon, Apple, Google and IBM all getting into the game.
And the connectivity is vast. Now, you can order an Uber with the help of ol’ Siri.
But no doubt, there is a barrier to voice taking over as an entire platform. See Mexican restaurant example above. But experts say speech recognition must achieve around 99 percent accuracy to become a more dominant platform. Currently, those percentages hover around 90, which isn’t terrible but is not what’s necessary.
But as we hone in on that 99 or even 100 number, what about companies that aren’t in on the voice game? Analysts say they’re at risk of being left in the voice dust. Nobody wants that, but rather, everybody wants a horse in the (voice) race.