Narrowing the Field in Voice Search

SRI International, an independent, nonprofit research institute, today announced that it’s been issued a U.S. patent for voice-to-database (V2DB) technology that allows users to vocally access in a single step information in multimillion-item databases.  

Doug Bercow, director of business development in SRI’s information and computing sciences division, believes V2DB is "more of an invention than a technology, and the invention provides a methodology to request specific items from a very large database with one utterance." He compares the development to the one-click ordering option on Amazon.com. When users search by voice through a space, they have to progressively narrow down their search criteria through iterations. "That’s where the item entry world is today," Bercow says, "but that’s not where we want it to end up."  

The use for this technology is great. Currently speech-enabled GPS systems, for instance, are limited due to the large number of street addresses. V2DB seeks to make that search space more manageable. Users, instead of having to narrow their search criteria by state and city, can instead state where they are currently and where they want to go, just as one might interact with a taxi driver.  

SRI however functions primarily in the invention arena as opposed to the realization of products. The patent for V2DB will be added to a patent portfolio of issued and pending patents for speech recognition technology that includes 87 other claims. "We’re in a campaign to package this patent portfolio to be licensed or acquired by companies who deploy these types of speech recognition companies," Bercow says.  These varied companies include the directory assistance industry, wireless network operators, and mobile application operators.      

Ultimately, Bercow believes SRI’s contributions should slot into existing core speech engines.  It’s particularly important as these technologies become  socially accepted, but also necessary to harness the world’s increasing volumes of searchable content. "We’re not inventing a new way of doing speech recognition," Bercow says. "It seems like it’s the kind of interface that will require some patience on the part of the industry as users begin to accept it, and it will also require some innovation and creativity by the speech industry to continue to put new kinds of applications and services out there."