        | |
|
TECHNOLOGY The Speak With Me architecture, which stems from more than 10 years of research at Carnegie Mellon University, was created with the following goals for manufacturers: • Easy development via rapid prototyping to help speed products to the market • Provide ability to customize product offerings to fit their consumer's needs • Ease of integration to recognizers and other applications to save costs on research and design and to offer a differentiated product • Provide a technology platform that offers high levels of accuracy and reliability What makes the Speak With Me architecture unique is that we approach the lack of accuracy in today’s speech technology as a “search” problem, rather than a speech recognition problem. We apply the latest natural language search techniques to determine the “context” and “meaning” a user evokes through their natural way of speaking via a real time context analyzer. The technology that processes this search is the Dialogue Manager. The Dialogue Manager’s processing of the user’s sentence structure, prior to invoking the services of a Speech Recognizer, is what enables Speak With Me to increase the accuracy of speech recognition from 70 percent accurate for a 700 word Dictionary (using IBM Embedded ViaVoice) to 99.6 percent for a 30,000 word Dictionary and 99.2 percent accurate for a 90,000 word Dictionary. Our platform has the following key attributes:
One of our goals was to create a platform for rapid prototyping. To achieve this goal, we developed a unique descriptive language application that allows manufacturers to customize the task of the Dialogue Manager as well as the back-end application. Using our descriptive language application, manufacturer’s can create custom applications within weeks rather than months. The ability of rapid prototyping and the accuracy of the Dialogue Manager provides the following key benefits:
|
 DEMOS Watch Now! |
| |||
| |