Database construction and analysis of user speech with real environment spoken guidance systems
Hiromichi Kawanami, Manabu Kida, Tobias Cincarek, Kiyohiro Shikano · The Journal of the Acoustical Society of America · 2006
A spoken guidance information system ‘‘Takemaru-kun’’ has developed and been operated in a public city center in Ikoma city, Nara, Japan since November 2002. The system employs the example-based response selection strategy and answers to users’ questions about the hall facility, the city, sightseeing, etc. with CG agent animation. Following the success of the system, two new spoken systems ‘‘Kita-chan’’ and ‘‘Kita-chan robot’’ were developed and were set on a railway station since March 2006. ‘‘Kita-chan robot’’ is being operated with a robotlike agent. All user speech of the systems is recorded and two year’s data of ‘‘Takemaru-kun’’ and one month’s data of ‘‘Kita-chan’’ systems are manually transcribed and labeled by age and gender. Proper answers are also given to each utterance. In this paper, details of the three user speech databases are described and analytical results concerning numbers of inputs and response accuracy and age group and gender are reported.