VEDAMA-spoken dialogue system for automatic correspondence

This paper presents the spoken dialogue system used for automatic correspondence. The system offers capabilities to voice command all the functions available in the system. This is the most natural way of man-machine communication. Architecture of the system is modular. It consists of four major modules: graphic interface, recognizer based on key-word-spotting, dialogue manager and word-based speech synthesizer. The speech recognizing engine was built using speech recognition and development toolkit, SPREAD. The system is developed for UNIX and Windows NT platforms. The dialogue manager implements four different levels of dialogue. The most complex level is for the owners and less complex for the others (friends, acquaintances of friends and for all other users). The caller can ask to listen to, record or deleting voice messages etc. The system itself is a half-duplex system and uses system driven interaction. In the paper general and functional structure of the system are discussed in more detail.