Google's Phoneme Bank
turbotodd 100000388Y Visits (2138)
That's the first thing you hear when you call into Google's voice search portal number, 1-800-GOOG-411.
But after hearing Google senior VP Marissa Mayer discussing its voice portal in more detail at last week's Searchnomics event in Santa Clara, and now after reading coverage of Google's acquisition of Grand Central overnight, a more comprehensive and strategic picture is beginning to emerge.
Google needs a few million of us to make contributions to its phoneme bank.
"Calls recorded for quality."
Uh-huh....and, phoneme analysis.
You know, those bits of language that constitute the theoretical representation of a sound which are critical building blocks for any good speech recognition capability.
And, more oxygen needed for the sprint to the mobile voice search recognition and results finish line against Microsoft.
Take a brief moment and flash back to Microsoft's acquisition of TellMe earlier this year, TellMe being a pioneer in 1-800 voice driven assistance and speech-to-Web capability.
As mentioned in a Reuters article at the time, "TellMe brings to Microsoft the world's largest database of voice-recognition data."
Realizing the import of that move (Microsoft using TellMe to drive its own voice and Web search integration), Google needed to exercise its own vocal chords and fast. Mayer admitted as much last week in Santa Clara.
What better way to collect even more phonemes to improve the algorithm than to buy a company (Grand Central) that has the ability record your phone calls and link you to a specific phone number!
Which brings us to the issue of privacy in the voice realm.
"If you do not have caller id blocked, we collect and store the number of your phone, along with the time of your call, each time you use the Google Voice Local Search service. We may use this phone number to distinguish you from other users, and ultimately, to personalize the service to you."
But there's more:
We also collect and store a copy of the voice commands you make to the service, so we can audit, evaluate, and improve the voice recognition capabilities of the service.
Calls recorded for quality and improved phoneme acquisition and analysis. The more phonemes, the better the GOOG 411 algorithm. Because the better the system gets at analysis, the higher the voice recognition rate.
Also, Google will now have the capability to analyze your voice print and potentially segment yours' from others.
Everybody wins. Users...advertisers and marketers (particularly those with a local bent).
But especially Google.
Because before too long, it will be Google which has the world's largest voice print database.
And then, the last mile of the Google personalized marketing juggernaut loop is closed.
Google will then have your single phone number (caller id recognizable), cookie, and voice print, along with a history of all your search queries and clickstream.
Throw in a few bat's heads and some mistletoe, and pretty soon you've got an extremely potent witches brew of information that makes marketers and advertisers salivate, and privacy advocates running for the Orwellian exits.
"Calls recorded for quality."