Speech recognition in the context of two-way immediate person-machine interaction

Speech recognition is too often considered for itself ; even the use of pragmatic constraints do not necessarily imply any two-way dialogue between speaker and machine. The speech feedback, if it comes within a few seconds, permits a true person-machine interaction. Such a dialogue process is described, using the speech understanding system ESOPE φ and the text-to-speech synthesizer ICOPHONE. It simulates a telephone operator. Other dialogue experiments have been carried out using isolated word recognition and different tasks. All these experiments show some peculiarities of person-machine speech communication, and give some guidelines for the definition of a further system.