A02 Communication understanding and generation research group

Research on the integration of understanding and generation modules for dialogue systems

Principal Investigator
Ryuichiro HIGASHINAKA (NTT Communication Science Laboratories, Senior Distinguished Researcher)
Special field : Dialogue system
Role : Management of group's research
Kazunori KOMATANI (Osaka University, Professor)
Special field : Speech dialogue system
Role : Research on speech and multimodal processing
Yusuke MIYAO (University of Tokyo, Professor)
Special field : Language processing
Role : Research on language processing
Michimasa INABA (The University of Electro-Communications, Associate Professor)
Special field : Dialogue system
Role : Research on dialogue system
Takashi MINATO (Advanced Telecommunications Research Institute International, Research Scientist)
Special field : Conversational robot
Role : Construction of robot system
Yasushi MAKIHARA (Osaka University, Associate Professor)
Special field : Image recognition
Role : Construction of sensor network system
Research Collaborator
Akinobu LEE (Nagoya Institute of Technology, Professor)
Special field : Voice recognition
Role : Research on speech recognition
Yasuhiro MINAMI (The University of Electro-Communications, Professor)
Special field : Voice recognition
Role : Research on speech recognition
Ryo ISHII (NTT Media Intelligence Laboratories, Senior Research Engineer)
Special field : Multimodal processing
Role : Research on multimodal processing

The Communication Understanding and Generation Research Group aims to challenge the architecture of current dialogue systems and to develop core technology for integrating understanding and generation modules so that the utility of a system can be maximized. Specifically, the following four issues will be addressed.

  1. Integrating speech and multimodal processing modules into other modules
  2. Integrating language understanding / language generation modules into other modules
  3. Establishing multi-module parameter optimization techniques based on the utility of dialogue
  4. Designing protocols for inter-module communication, constructing dialogue systems based on these protocols, and demonstrating the effectiveness of such systems

By focusing on these issues, the modules can work together to improve the performance of each module. In addition, the performance of each module can be adjusted so as to maximize the utility of the whole dialogue. Protocols and design guidelines for integrating modules will be made open to the public, and the effectiveness of systems built based on those protocols and guidelines will be verified through social implementation.