Broad Area Colloquium For AI-Geometry-Graphics-Robotics-Vision
(CS 528)


Knowledge Systems and Project Halo

Bruce Porter, University of Texas, Austin
April 11, 2005, 4:15PM
TCSeq 201
http://graphics.stanford.edu/ba-colloquium/

Abstract

Knowledge Systems are computer programs that use broad knowledge about a domain to answer a wide range of novel questions with coherent explanations. Our goal is to develop methods by which scientists, with no training in knowledge engineering, can build large knowledge systems. This talk will describe our approach of building knowledge systems compositionally by instantiating and assembling representational components of general actions and entities.

Project Halo is a challenge problem designed to gauge the current state of knowledge systems. Vulcan, Inc. hired three teams to build knowledge systems for Advanced Placement (AP) chemistry. (AP is an exam for gaining course credit at US universities.) The talk will also discuss the novel experimental design, as well as the architecture of our winning submission.

About the Speaker

Bruce Porter is a Professor in the Department of Computer Science, and the Director of the AI Lab, at the University of Texas in Austin. Professor Porter completed his PhD in machine learning in 1984 at the University of California, Irvine. He earned a Presidential Young Investigator award from the National Science Foundation in 1987. With the goal of building large, useful knowledge systems, his research focuses on knowledge representation and reasoning, knowledge acquisition, and explanation generation. In 1997, he and colleague Peter Clark wrote a research paper laying out their vision for building knowledge systems compositionally. The paper was selected for a Best Paper Award by the National Conference on Artificial Intelligence.


Contact: bac-coordinators@cs.stanford.edu

Back to the Colloquium Page