Telephone dialogue data base of JST/CREST Expressive Speech Processing Project

JST/CREST Expressive Speech Processing Project has been constructing a large-scale corpus of spontaneous Japanese dialogue under the fund of Japan Science and Technology Corporation. The corpus is designed to contain about 1000 hours of digitized dialogue speech to realize near future friendly computers which can speak like a human being. This paper reports the objective and the data collection method of the telephone dialogue database as a part of this corpus.