5th International Conference on Spoken Language Processing
This paper proposes a mechanism that contributes to the implementation of a spoken dialogue system with which a user can communicate effortlessly. In a dialogue, exchanges between participants promote the establishment of shared information and this leads to effortless communication. This is called "dialogue coordination". In particular, revealing the respondent's internal state, such as through nodding and back-channel feedback, promotes the establishment of shared information. This is called "manifestation", which is one aspect of coordinative behavior, and a mechanism for handling manifestation is introduced. In a human-human dialogue, the listener's manifestative behavior often occurs during a speaker's utterance. However, systems using conventional speech recognition technologies cannot respond during the speaker's utterance. In order to solve this problem, the proposed mechanism, ISTAR protocol transmission, utilizes the intermediate speech recognition results without waiting for the end of the speaker's utterance. This realizes a system with flexible manifestative behavior.
Movie Clip - This design of the manifestative behavior was implemented, and the dialogue for this implementation is in the movie clip [.MOV file].
Bibliographic reference. Hirasawa, Jun-ichi / Miyazaki, Noboru / Nakano, Mikio / Kawabata, Takeshi (1998): "Implementation of coordinative nodding behavior on spoken dialogue systems", In ICSLP-1998, paper 0158.