MPML for Describing Multimodal Contents with Character Agents
Speaker: Professor Mitsuru Ishizuka, School of Information Science and Technology, the University of Tokyo
When: 2004-10-07 15:00:00
Venue: G23_2.07a Gold Coast campus, Griffith University
Host: Prof Abdul Sattar
Abstract:Embodied lifelike agents allow a new style of multimodal
presentations and interactions. Since they are performed by the
agents instead of human presenters, people can audit and enjoy them
"anytime" and "anyplace" through the network. In addition, it would
be desirable if "anyone" can write attractive and affective
multimodal contents easily, like many people write their own Web
contents by themselves using HTML or available graphical editors.
In this talk, I will show the results obtained in our recent project
with some demonstrations. The first focus of the project is the
development of a family of XML-based content description languages
called MPML for controlling lifelike agents in various environments
including the Web, 3D space (i.e., VRML space) and mobile phones.
The main mode of MPML usage is scripting the verbal and nonverbal
behaviors of the character agents, because scripting rather than
generating behaviors autonomously is suitable at present as a
practical tool. While keeping this main mode, the second focus of
the project has been a moderate introduction of functionalities for
(semi-) autonomous behaviors so as to facilitate the content
scripting. MPML provides a function for supporting the emotional
manifestation of the agents. Emotion, together with personality, is
an important factor for improving the lifelikeness and believability
of the agents, and thus making the presentation attractive,
impressive and memorable. Another issue related to the autonomy of
the agent is conversational flexibility. For this issue, we have
enabled a combination of a chatterbot and an MPML-scripted
presentation, where a smooth transition between the two
presentation/dialogue patterns is achieved.
Biography:Mitsuru Ishizuka has been a professor at the Dept. of Information
and Communication Eng., School of Information Science and
Technology, the University of Tokyo since 1992. Prior to this
position, he worked at NTT Yokosuka Lab., Purdue Univ., and the
Institute of Industrial Science (the Univ. of Tokyo). He earned his
B.S., M.S. and Ph.D. in electronic engineering from the Univ. of
Tokyo. His research interests include artificial intelligence, WWW
intelligence, and multimodal lifelike agents. He was the program
co-chair of PRICAI-02 with Prof. Abdul Sattar. Presently, he is the
president of JSAI (The Japanese Soc. for AI). His homepage is at
http://www.miv.t.u-tokyo.ac.jp/~ishizuka/eng.html.
Type: Colloquium Seminar
Contact:Prof Abdul Sattar, seminar host (n.dunstan@griffith.edu.au)
or Guido Governatori (ITEE seminar co-ordinator)
(guido@itee.uq.edu.au)