Yigal Arens, Craig Knoblock, Wei-Min Shen
USC/Information Sciences Institute
arens@isi.edu, knoblock@isi.edu, shen@isi.edu
Most tasks performed by users of complex information systems involve interaction with multiple knowledge- and data- bases. Examples can be found in the areas of analysis, resource planning and briefing applications. Retrieval of desired information dispersed in multiple sources requires general familiarity with their contents and structure, with their query languages, with their location on existing networks, and more. The user must break down a given retrieval task into a sequence of actual queries to databases and/or knowledge bases, and must handle the temporary storing and possible transformation of intermediate results - all this while satisfying constraints on reliability of the results and the cost of the retrieval process. With a large number of information sources, it is difficult to find individuals who possess the required knowledge, and automation becomes a necessity.
SIMS presents an elegant solution to the problem described above: the creation of a knowledge server - an information mediator. A user, or an application, queries the mediator in a manner that is independent of the distribution of information over various sources, of the various query languages, the location of data- and knowledge- bases, etc. It is SIMS' task to determine how to obtain the desired information, which data sources to use, how and where to temporarily store and manipulate intermediate data, and how to maintain a satisfactory level of efficiency.
Paul Rosenbloom, Milind Tambe
USC/Information Sciences Institute
rosenblo@isi.edu, tambe@isi.edu
Our research effort is aimed at developing human-like, intelligent agents for large-scale, interactive simulation ("virtual reality") environments. Such "virtual humans" promise a number of benefits for the virtual worlds, with applications in education, manufacturing, training, and entertainment. To begin this effort, we have currently focused on building intelligent pilot agents to control simulated military aircraft in simulated battlefield environments. The demo will show intelligent pilots flying fighter jets and helicopters, and engaging in simulated air-to-air as well as air-to-ground combat.
Kevin Knight
USC/Informationa Sciences Institute
knight@isi.edu
We will demonstrate a system that translates newspaper articles from Japanese to English, and we will describe how we evaluate the quality of our output.
Lewis Johnson
USC/Information Sciences Institute
johnson@isi.edu
This demonstration will present the Debrief system, which provides intelligent agents implemented in Soar with the ability to describe and explain their actions. Explanation capabilities are necessary in order to convince users that intelligent agents can safely be assigned tasks to perform. The demonstration will Debrief explaining the actions and decisions of air-to-air combat agents in a distributed battlefield simulation. Debrief's explanations employ a combination of diagrams and natural language text. The text is mouse-sensitive, allowing users to select fragments of text and ask follow-on questions.
Lewis Johnson
USC/Information Sciences Institute
johnson@isi.edu
This project is building a dynamic software documentation environment, which presents focused descriptions of software systems sensitive to the reader's task context and level of expertise. These presentations are generated using a customized World Wide Web server that extracts information from a design repository implemented in Refine and Lisp, and then composes descriptions in HTML based on this information. This application illustrates how the Web can be used as a medium for complex interactive applications.
Yolanda Gil, Bing Leng, Bill Swartout, Marcelo Tallis
USC/Information Sciences Institute
gil@isi.edu, leng@isi.edu, swartout@isi.edu, tallis@isi.edu
EXPECT is an environment that supports users in building, maintaining, and extending knowledge-based systems. In order to make these systems more accessible to end users, we are building tools that allow them to understand what a system is doing and to update and fix the system's knowledge through interfaces that communicate in meaningful domain terms. EXPECT includes an explanation facility that generates natural language descriptions of the system's knowledge. We will demonstrate how users can examine and extend the system's knowledge bases through EXPECT's knowledge acquisition interface.
Lorna Zorman
USC/Information Sciences Institute &
Computer Science Department
lorna@isi.edu
Requirements envisaging is the process of transforming peoples' informal notions of what is desired into a precise description. Often, the form of description that people find easy to use is scenarios: partial descriptions of system and environment behavior arising in restricted situations. I will be demonstrating a formal representation for scenarios embodied in a direct-manipulation interface that allows domain and software experts to create scenarios.
Allen Munro, Mark C. Johnson, Quentin A. Pizzini
David S. Surmon, Douglas M. Towne, James L. Wogulis
Behavioral Technology Laboratories
University of Southern California
250 No. Harbor Drive, Suite 309
Redondo Beach, CA 90277
RIDES is an interactive application for composing and delivering graphical simulations and simulation-based training. Authors use this tool to draw graphical objects and to specify their behaviors with respect to each other and with respect to user actions. This process creates editable interactive graphical simulations. Authors can then create procedural training vignettes by carrying out sequences of actions in the simulation. Later, trainees will be required to carry out the same procedures during training.
The RIDES application has a number of interesting user interface features, three of which are emphasized in this demonstration. These are: (1) authoring procedures by direct manipulation, (2) specifying object behaviors with constraints and with events, and (3) the primacy of reference (rather than of name) in behavior specification.
Stephen Casner
USC/Information Sciences Institute
casner@isi.edu
Synopsis:
Worldwide teleconferencing with 10's or 100's of participants is now possible over the Internet MBONE using audio, video and shared whiteboard tools available for free.
Abstract:
During the past two years, there has been dramatic growth in the development and use of real-time audio and video teleconferencing over the Internet. This growth began with the transmission of live audio and video from meetings of the Internet Engineering Task Force (IETF) to a remote audience that has expanded from 20 to over 600, and led to the establishment of the Multicast Backbone (MBONE). The MBONE is a virtual network overlaid on portions of the physical Internet to provide IP multicast connectivity among the participating sites.
For this demonstration, the software tools for teleconference session control and for audio, video, image and shared whiteboard media will be shown in action on workstations at ISI. We will try to teleconference with other folks around the world who might be available.
These tools have been developed by several organizations and have become known as "the MBONE tools". Attendees are encouraged to learn how their home institutions can connect to the MBONE and join in this experiment.
Ewald Salcher
salcher@isi.edu
Visiting Ph.D. Student from
Graz University of Technology Institute for Computer Graphics
Muenzgrabenstrasse 8010 Graz, Austria
Typically an interactive application consists of a data model and a graphical user interface which presents this model to the user. Much effort goes to the mapping of high level application objects to low level interaction techniques. This approach tries to abstract modeling in terms of semantic concepts and to generate the data model and the interface model directly from a semantic model.