Google's latest granted patent (
US Patent No. 7,027,987) describes searching by voice. The basic methodology is as follows:
- Voice input of a user is captured (say from a microphone or during a phone call).
- Probable words are extracted from the voice input using speech recognition.
- The probable words and variants are searched for in Google.
- Probable results are presented to the user.
The speech recognition can be customized based on the locale and/or accents. The results can also be categorized based on the probable words.
I'm a little confused about how Google will implement voice searching. In the presence of a keyboard, entering information would be more precise (albeit a bit slower) as compared to voice input. The accuracy of the results would, of course, be much better. I have yet to use a speech recognition engine that can construct a sentence based on what I say.
The obvious implementation appears to be in mobile devices such as cellphones, PDAs and the recently launched
Ultra Mobile PC (UMPC), which still offer clumsy keypads. Further, this service would be most useful for single answer searches - such as
flight tracking,
FedEx package tracking,
currency conversion, etc.
Maybe someday, Google will be able to tell you
where your car keys are. All you will have to do is to ask the right question.
Point to be noted:
Monica Henzinger, one of the inventors, looks surprisingly like one of my friends.
Further reading:
US Patent No. 7,027,987