In this paper we describe a speech dialogue system for the field of directory assistance. The demonstration task is a directory with 5000 entries, currently. The speech technology employed, however, is feasible for databases of realistic size, too. The system uses keyword spotting, alphabet recognition, and speech synthesis. Recognizer and syntheziser are based on a German pronunciation lexicon. The combination of keyword spotting and alphabet recognition allows, to enlarge the vocabulary to be recognized significantly. The overall system relies on a general architecture for voice dialogues, which was designed to be open for other base components and further development in speech technology. The dialogue software exhibits some general design rules for speech dialogue systems.
Bibliographic reference. Kaspar, B. / Fries, G. / Schuhmacher, K. / Wirth, Antje (1995): "Faust - a directory assistance demonstrator", In EUROSPEECH-1995, 1161-1164.