IJCAI-97 Workshop on
Intelligent Multimodal Systems


Nagoya Congress Center, Room 223

Nagoya, Japan, August 24 (Sun), 1997

in connection with

15th Int'l Joint Conf. on Artificial Intelligence


[Updated August 19, 1997]

Scope
Call for Papers [closed]
Call for Participation
Workshop Format
Program with links of contributed papers
Another Program Page (Nagao's On-line Proceedings)
Organizers and Program Committee

　Invited Talks

Knowledge-Based 3D Graphics
by Prof. Steven Feiner (Columbia Univ.)

As 3D computer graphics hardware and software grow in power and popularity, potential users are increasingly confronted with the daunting task of using them effectively. Making the decisions that result in effective graphics requires expertise in visual design, effort, and time, all of which are scarce commodities. I will discuss research that addresses these issues by exploring the possibility for automated knowledge-based design of interactive 3D graphics and virtual worlds. Examples will be given from domains ranging from explanations of physical tasks to visualizations of abstract data.

Multimodal Anthropomorphic Interface
by Dr. Yoichi Takebayashi (Toshiba R&D Center)

The Graphical User Interface (GUI) first prompted the spread of personal computers since its "See and point" interaction was simple and easy to use. Then, the recent advances in computing and networking environments enabled general users to handle massive digital data, thereby influencing our everyday life. This implies that our research should concentrate on what support computers can give to humans and society by adding perceptual capabilities such as speech and vision to the human-computer interface.
The problem of the so-called "information flood" can be solved by the multimodal anthropomorphic agent which understands users' intention and conditions as well as the contents of information through "Ask and tell" interactions. To illustrate my point, I will first demonstrate our three systems, a spontaneous speech dialogue system TOSBURG II and two multimodal interface agents for enhancing knowledge sharing, and then address some research issues for achieving highly practical systems.

For more information:

Mitsuru Ishizuka
Dept. of Information & Commun. Eng., School of Eng., Univ. of Tokyo
7-3-1, Hongo, Bunkyo-ku, Tokyo, 113, Japan
Phone: +81-3-3812-2111 ext.6347 Fax: +81-3-5802-2990
Email: ishizuka@miv.t.u-tokyo.ac.jp

or, See IJCAI-97 .

SCOPE

The use of multimodality holds promise to enhance human-computer interaction. Since human perception is based largely upon multimodality, it is an important factor in the realization of human-friendly interfaces.

Multimedia is a closely related term to the multimodality; it can be said that the 'multimodality' is used mainly to refer to multiple interface channels related to human cognitive process, whereas the 'multimedia' is used widely to refer to various computer-related technologies for building multimodal/multimedia systems. Furthermore, recent advances in virtual reality and devices promise interesting possibilities of human perception unified with synthesized environments.

In addition to media technologies such as image, graphics, sound, voice, and natural language processing, etc., artificial intelligence (AI) technologies are now expected to enhance the flexibility and friendliness of the multimodal interfaces and to provide intelligent support, thus broadening the application spectrum of human-computer interaction. Recent progress of media and computer technologies allows us to build interactive multimodal information systems, and early implementations promise exciting possibilities. However, systems with intelligent functions are few to date. At present, we don't know exactly which areas of intelligent multimodal systems will be the most promising and we don't know yet what are the key factors to practical success. Since many technologies have to be integrated into a multimodal system, it is important to clarify research foci.

To this end, this workshop is being organized in order to exchange ideas obtained from various experimental studies on this field. Rather than discussing component technologies, we wish to emphasize the practical or experimental implementations of intelligent multimodal systems.

New multimedia technologies such as intelligent multimedia interfaces are included in the topics of the workshop. Also, systems in virtual reality environments can be a topic. In response to emerging new network-based information environments, papers on multimodal systems based on the Internet are welcome. The topics of the workshop include (but are not limited to);

WORKSHOP FORMAT

The workshop consists of
  1. presentation of paper summaries (10 min. each),
  2. two invited talks, and
  3. open discussion on selected issues.
Emphasis will be placed on discussions on the following selected issues regarding Intelligent Multimodal (IMM) Systems.
  1. taxonomy of modalities in IMM systems,
  2. where is the intelligence in IMM systems,
  3. promise/problems of anthropomorphic interfaces,
  4. major obstacles to industrial take up of IMM systems,
  5. mechanisms to improve modularization and re-use of IMM systems,
  6. research agenda for IMM systems.
To facilitate the discussion, the authors of contributed papers and participants of the workshop are requested to take the following actions before the workshop.
  1. The author of each contributed paper prepares the PS file and HTML file (Web page) of his/her paper, which is linked from the workshop Web page (http://www.miv.t.u-tokyo.ac.jp/ijcai97-IMS/).

    {K. Nagao (nagao@stereo.ai.uiuc.edu) will help authors to prepare the HTML version of papers. If necessary, please send a (LaTeX or plain) text file and linked GIF or JPEG files to him.}

  2. The PC members and participants of the workshop develop some themes/questions about the papers and their approaches to the main issues of IMM systems. These are to be placed on the workshop Web page.

  3. The authors are asked to write answers to the questions in the light of their papers. These answers will be available on the Web, if possible, before the workshop.

  4. All the participants and the PC members also write their comments on the selected issues (1 - 6) listed above, before the workshop. We wish to draw some consensus on these issues at the workshop.
{Comments/questions/answers are to be directed to M. Ishizuka (ishizuka@miv.t.u-tokyo.ac.jp) as well as to K. Nagao (nagao@stereo.ai.uiuc.edu) .}

PROGRAM

August 24 (Sun), 1997

Opening session [9:30 - 9:40]

Session 1: Mutlimodal Functions for Robots [9:40 - 10:10]