ESCA Workshop on Spoken Dialogue Systems Vigsø, Denmark May 30 - June 2, 1995

A Prototype of an Address Input System with Speech Recognition

Kazuhiro Arai, Osamu Yoshioka, Shigeki Sagayama, Noboru Sugamura

NTT Human Interface Laboratories, Kanagawa, Japan

It is still currently very difficult to obtain a high level of performance for spontaneous speech recognition, even though there have been many advances toward the ultimate goal of a speech dialog system which is to achieve natural interaction with the user. Isolated word recognition has presented itself as a prospective candidate for obtaining a higher level of performance in practical use. Therefore, one reasonable way for developing a useful system is to employ isolated word recognition. We have developed a system for address input that integrates isolated word recognition and conventional input methods. The input time was measured for 25 adult novice users, with 22,860 trials in total to test if the speech recognition function is useful. The results revealed that speech recognition reduces the time required for input when the input volume is large.

Full Paper

Bibliographic reference.  Arai, Kazuhiro / Yoshioka, Osamu / Sagayama, Shigeki / Sugamura, Noboru (1995): "A prototype of an address input system with speech recognition", In SDS-1995, 213-216.