Voice Recognition Evolving From Many Development Efforts

 
 
By Robert J. Mullins  |  Posted 2012-06-22 Email Print this article Print
 
 
 
 
 
 
 


Coincidentally, perhaps, Google in March hired a former director at DARPA, Regina Dugan, but it is not clear what her role will be or whether she worked on the CALO project.

While Microsoft and Apple have their voice command innovations, Google has long been involved in delivering voice-recognition technology, too, even prior to the launch of its Android mobile OS, said William Stofega, an IDC program director in Mobile Device Technology and Trends research.

Stofega was a longtime fan of Google 411, the directory-assistance service that was based on voice recognition. People would call G-O-O-G-4-1-1 on their mobile phones and get a phone number for a person or business and dial through to that number for free. Google launched it in 2007 but dropped it in 2010.

€œAt first, there were troubles, but then you started to see it start to learn and get better and better,€ Stofega said. €œBy the time they pulled it off the market, it was great; it was fantastic and I miss it.€

Other voice-recognition technology leaders include Nuance Communications, famous for the Dragon line of voice-to-text dictation software. Stofega says Nuance technology is behind the Siri service, although Apple also adds proprietary artificial intelligence technology so that the actress Deschanel can tell Siri, €œLet€™s get tomato soup delivered€ in one TV ad, and Siri can respond with a list of area restaurants that deliver tomato soup. The Siri technology comes from an Apple acquisition in 2010.

Microsoft€™s voice-recognition smarts come from its 2007 acquisition of Tellme Networks, a voice-recognition company that is operated as a wholly owned subsidiary of Microsoft.

Stofega is skeptical that any voice service can be as accurate as their portrayals in TV ads€”as they can fail to recognize heavy accents and be affected by ambient background noise€”but says the technology keeps improving.

Analyst Ellison says modern voice recognition in systems like Siri or Audible depends on three key factors. First is the quality of voice recognition, the ability for the software to understand the words being said.

Second is determining the meaning of the words, he said. On interactive voice-recognition systems (IVR) used in a customer service setting, there are prompts like €œSay €˜Balance€™€ if someone is calling their bank to see how much money is in their checking account. As IVR€™s evolved, however, the systems have learned to understand the meaning of a string of words in context, allowing a customer to instead ask €œHow much money is in my checking account?€

Third, Ellison explained, is the integration of the voice-recognition technology and the application with the database of information available to the Siri or Audible service.

€œThe [Siri] engine needs to understand and format this data in a way that another application provider can use. This service needs to be able to meta-tag their data appropriately to provide the appropriate responses,€ Ellison said.

A smartphone personal assistant, as the technology evolves, can be a real aid to public safety, said analyst Stofega. Traffic accidents caused by distracted driving can occur because drivers are fumbling with their touch-screen phones to find information or, worse, sending text messages. As voice services evolve, drivers will be able to dictate texts and ask for directions to that place that serves that great tomato soup.



 
 
 
 
Robert Mullins is a freelance writer for eWEEK who has covered the technology industry in Silicon Valley for more than a decade. He has written for several tech publications including Network Computing, Information Week, Network World and various TechTarget titles. Mullins also served as a correspondent in the San Francisco Bureau of IDG News Service and, before that, covered technology news for the Silicon Valley/San Jose Business Journal. Back in his home state of Wisconsin, Robert worked as the news director for NPR stations in Milwaukee and LaCrosse in the 1980s.
 
 
 
 
 
 
 

Submit a Comment

Loading Comments...

 
Manage your Newsletters: Login   Register My Newsletters























 
 
 
 
 
 
 
 
 
 
 
Rocket Fuel