iPhone Home, Hands Free
With a growing number of speech-enabled applications for the iPhone, owners of Apple’s popular smartphones no longer need to feel burdened with the oppressive yoke of a swelling list of friends, co-workers, family members, and other contacts. Speech search and recognition technologies provider Melodis earlier this week released Voice Dialer, a freeware application that eliminates the need to scroll through a long list of names, addresses, and phone numbers to make a call.
The Voice Dialer application is a response to what Melodis sees as a limitation with the iPhone’s innate touch-based contact interface, which can become slow and difficult as the list grows too large. Instead, Voice Dialer allows users to interface with their iPhones vocally.
Once installed, the application automatically indexes all the contact information already stored on the iPhone without the user having to re-enter it. To begin using Voice Dialer, the user has only to hold down the application button and speak her commands. If she wanted to call her friend “Cathy Hebert,” for instance, she would say, Call Cathy Hebert to get Voice Dialer dialing. Additionally, the software can handle multiple matches. If, for instance, there were also a “Kathy Hebert” listed, the command Call Cathy Hebert would bring up both Cathy and Kathy and let the user select the appropriate one. Moreover, just saying Call Cathy would bring up every Cathy, Kathy, Cathie, Kathie, and similar homophonically named person in the iPhone’s memory.
The software allows the user to get even more specific with call commands, too. Our user could call Cathy at work by saying Call Cathy Hebert at XYZ Corporation, or, use the same command to distinguish between multiple listings with the same name. She could also get a directory of everyone she knows at XYZ by just saying XYZ Corporation. Because Voice Dialer indexes each contact entry in its entirety, any aspect of the entry can be searched for vocally.
The application builds on Melodis’ existing Crystal Engine, a speech search engine that accepts voice as input directly rather than relying on traditional speech-to-text conversion first. According to Keyvan Mohajer, president and CEO at Melodis, this helps the engine achieve a very high degree of accuracy.
“A lot of companies are trying to get into this space, but almost all of them license the core engine from another provider,” he explains.
These engines are usually built for transcription and run against limitations when applied to search solutions, Mohajer adds. “If there is a mistake in the voice-text transcription, that propagates to the next stage and damages the search,” he says.
To increase accuracy in this scenario, one sacrifices speed, but, by foregoing transcription altogether, the Crystal Engine can be fast without losing accuracy, Mohajer explains
The software is available for free and will be ad-supported in the future, pending negotiations. The ads will come in the form of banners at the bottom of the iPhone screen and may be targeted specifically to the user.