Speech Offers a New Pathway to Directory Assistance

Voice-activated directory assistance provider Dial Directions and Jingle Networks, operator of free voice search service 1-800-FREE411, announced today a strategic partnership. Under the partnership, Jingle will leverage Dial Directions’ technology, allowing callers to 1-800-FREE411 to receive SMS directions to their spoken destination.

To receive directions, callers speak their starting and ending destinations and receive turn-by-turn directions via text message. While the service is already available nationwide through Dial Directions own phone service (available by calling DIR-ECT-IONS or 347-328-4667), the Jingle service, available in the San Francisco Bay Area, Chicago, and Los Angeles by March 1, will expand nationally 30 days later.

Uptake of Dial Directions’ service should increase call rate and frequency to FREE411 and thus increase Jingle’s advertiser revenue, which Dial Directions and Jingle Networks will share.

Part of Jingle’s decision to add Dial Directions’ service is to stay competitive with Google’s free 411 service GOOG411. The partnership is a proactive step forward, according to Dial Directions CEO Amit Desai. He anticipates the partnership will be a long-lasting one "because we’re changing the expectations of DA [Directory Assistance]. [Jingle is] moving a step forward, but that means everyone else will want to keep up with the Joneses, so you can’t fall behind."

Dial Directions, which launched a beta version of its service in July 2007, was founded by a group of speech recognition experts and voice interface specialists. One of the major problems Dial Directions initially faced was developing a voice interface for frequent-usage systems. Earlier systems developed during the first few years of the millennium were, according to Desai, inadequate in terms of simplicity, accuracy, and speed. Thus, Dial Directions puts a lot of IP into its speech recognition engine.

"With frequent usage, you need an architecture that supports heavy personalization," Desai adds. "That means the system should know how many times the caller has seen this or that part of the interaction and adjust the interaction accordingly."

For example, caller history should tell the system when a user is contacting the system for the fifteenth time. In response, the prompts should be comparatively terse and quick. If however on the 15th call, the user accesses a new part of the system, the system should have the intelligence to revert to introductory prompts.

"We do a lot of tuning and personalization," Desai says. "Our service has been around in beta and we’re learning a lot based on how people are behaving towards the system. That data we’re transforming into analysis and using that to change the call flow, make it simpler, basically adapting to how people are using it."