Members
- Group1: Development of a CA which has a humanlike presence and a lifelikeness
- Group2: Research and development on unconstrained spoken dialogue
- Group3: Human level knowledge and concept acquisition
- Group4: Cooperative control of multiple CAs
- Group5: Development of CA platform
- Group6: Multidisciplinary investigation on how avatars and devices affect users
- Group7: Field experiments in the real world
- Group8: Avatar social ethics design
Group2: Research and development on unconstrained spoken dialogue Activities
Members
- Group leader
-
- Tatsuya KAWAHARA (Kyoto University)
Research and development for unconstrained remote spoken dialogue
- Tatsuya KAWAHARA (Kyoto University)
- Principal investigator
-
- Hiroshi SARUWATARI (The University of Tokyo)
Research and development for acoustic information processing and voice conversion - Ryuichiro HIGASHINAKA (Nagoya University)
Research and development for dialogue knowledge processing - Akinobu LEE (Nagoya Institute of Technology)
Research and development of CG-CA specific dialogues - Kazunori Komatani (Osaka University)
Research and development for robust spoken dialogue processing
- Hiroshi SARUWATARI (The University of Tokyo)
Overview
Our work will be on the development of the flexible dialogue technology which will enable operators using CAs to switch flexibly between two CA modes: the teleoperated and autonomous modes. This will realize a CA with a high degree of hospitality.
In order to achieve this goal, we first develop an autonomous CA that can talk with human-level hospitality, and then implement a semi-autonomous dialogue system that integrates the autonomous system with teleoperated one. The system includes functions such as attentive listening, basic questioning-answering, and casual chatting. For flexible mode switching, we will develop technologies for natural speech-recognition, -conversion, and -synthesis, especially using end-to-end deep learning. Our work will be conducted in close connection with other Groups.
Goal
We will develop "unconstrained dialogue technology" in which the CA flexibly switches between teleoperated dialogue and autonomous dialogue.
The hospitality of the autonomous dialogue system is evaluated with linguistic analysis and subjective evaluations using questionnaires. The integration of teleoperated and autonomous dialogue, attentive listening, question-answering and casual chatting are evaluated with methods and experts of psychology. Automatic speech recognition is evaluated with standard database and speech conversion and synthesis are evaluated with subjective evaluations.