Speech Technology Magazine

 

Star Performers: Microsoft Makes Mobile Moves with Cortana

By Leonard Klie - Posted Jul 29, 2014
Page1 of 1
Bookmark and Share

With Apple and Google battling for complete dominance of the smartphone market, Microsoft is often viewed as an outsider, but you wouldn't know it from the amount of resources it's pumping into improving the speed and accuracy of the speech recognition, speech-to-text, and voice search features on its Windows Phone mobile devices.

Microsoft's mobile team has been working with Microsoft's research division for more than a year to improve its technology using an advanced approach called deep neural networks, which allow technology systems to detect patterns akin to the way biological systems recognize patterns.

The work has already started to pay off. When a user starts a voice search on the latest release of Windows Phone, Bing returns results twice as fast as before, and speech recognition accuracy is increased by 15 percent.

In a recent blog post, senior Microsoft researcher Dong Yu calls the work with deep neural networks "just the first step in advancing the state of the art. Many difficult problems may be attacked under this framework, which might lead to even greater advances."

Furthering its push into the mobile space, Microsoft in April introduced Cortana, a voice-activated personal digital assistant for Windows Phone. Cortana is powered by Bing and can respond to conversational requests entered by voice or text input to search the Web, set alarms, manage calendars, make and receive calls, send and receive text messages, control settings and other apps on the phone, take notes, get driving directions, play music, issue reminders, and perform a host of other actions. The mobile application can respond with its own voice or text output, which is being outfitted with traits for 16 emotions.

Cortana will also be able to learn from and deliver more personalized search results based on personal information, including the user's likes and dislikes, interests, locations, and personal contacts, which Cortana collects and stores in her Notebook.

While comparisons to Siri are inevitable, Cortana separates herself with a capability called multistep search and the ability to plug into third-party apps. When asked to search for a restaurant, Cortana returns one result, based on the data she has in her Notebook and Yelp reviews. Once the restaurant listing is displayed, the user can say, "Get directions there," and Cortana, unlike Siri, understands that "there" refers to the restaurant she just pulled up. This multistep search creates more of a conversational feel than a series of disjointed questions and answers.

Mobile industry analyst Jeff Kagan isn't surprised by Microsoft's entry into the virtual assistant space. "Microsoft is finally joining the party," he says. "It started with Apple and then Google and now Microsoft. I don't think many thought that Siri was going to be duplicated, but apparently it's the new battleground."

How successful will Microsoft's entry be? Kagan says Cortana is "needed and wanted," but the world will have to wait and see just how good it is. "If it's as good as Microsoft says, this will be important for [its] growth going forward," he adds.

That's especially true if Microsoft follows through on stated plans to bring Cortana to laptops, PCs, and tablets running on the Windows operating system.


Page1 of 1