Press "Enter" to skip to content

Google, Microsoft look beyond mobile search for voice interaction

Organizing all of the world's information is no good unless that information can be accessed, and Google's recent move into free 411 searches shows that the company is serious about voice search. But Google's plans appear to extend far beyond finding local businesses. In a patent issued last year, the company outlined a full-blown natural language voice search platform, and Microsoft's own recent acquisition of TellMe indicates that a voice search war could be brewing. 苏州美睫美甲

The Google patent, issued in April 2006 and naming Sergey Brin as an inventor, describes a system for providing search results from voice queries, but it goes well beyond the system currently implemented as GOOG-411. In the patent, the inventors recognize the problems inherent in this sort of searching: lack of context and very short queries. "There is very little repetition in queries, providing little information that could be used to guide the speech recognizer," they note. "In other speech recognition applications, the recognizer can use context, such as a dialogue history, to set up certain expectations and guide the recognition. Voice search queries lack such context."

Nevertheless, Google believes that it can solve the problem, and the fruits of that effort appeared to be on display in GOOG-411, which can take moderately constrained input data (a business or business category, constrained by city) and return useful results. Though limited to business search at the moment, the system can handle a huge variety of accents and ways of requesting the same information, and pairing voice queries with the eventual search results is an excellent way to build up a database of useful voice snippets. This is exactly what Tim O'Reilly thinks is going on: Google is using an early implementation of the system to gather enough data to improve its voice recognition algorithms before a broader launch.

Voice search has been heating up in the last few months; one of Microsoft's largest buyouts this decade was made earlier this year, when the company acquired TellMe, which develops voice-recognition applications for use over the telephone. One of its most successful products is an automated 411 system; could Google's launch of a similar service less than a month after the Microsoft acquisition be pure coincidence? It certainly could, but it could just as easily be a show of strength from Google, which wants to make clear that Microsoft's expensive purchase can already be generally replicated by Google engineers.

Voice search, especially over mobile phones, stands to be a huge money-makerfor firms that can reliably automate the process of requesting information. Even directory assistance, which Google currently offers for free, is a cash cow, currently pulling in a buck a call or more. Late last year, the Kelsey Group estimated that directory assistance alone would service 6.3 billion requests in 2006—and that says nothing about demand for more general search services.

With that kind of potential revenue on the line, both Google and Microsoft hope to position themselves as leaders in the field. Both have been working with voice recognition for years; Google's patent discussed above was filed in 2001, and Microsoft's Richard Sprague told me at a recent conference that his company's voice recognition technology (now built into Vista) had been developed completely in-house over the last several years.

Nuance, another big player in the voice recognition market, won't offer a voice search tool itself, but already provides the tools to do so to third-party providers. And Yahoo, unlikely to abandon such a potentially lucrative market to its rivals, will no doubt make a voice search announcement of its own in the next six months.