A project of the Knowledge Systems Group, UT-Austin
The cornerstone of our approach to building Knowledge Systems is the assembly of knowledge bases from reusable components. Each component represents an entity, event, role or property. By instantiating and combining these components, complex representations can be built simply. For an overview of the approach, see our AAAI-97 paper ("Building Concept Representations from Reusable Components") and our K-Cap '01 paper ("A Library of Generic Concepts for Composing Knowledge Bases").Our inventory of general, domain-independent concepts is inspired by existing ontological and lexical resources, including:
In the end, our component library will be well integrated and cross-indexed with many other resources, and it will add a level of detail and computational support that the others omit. We envision that the component library will be a valuable resource for projects in knowledge engineering, natural-language processing, information retrieval, and machine learning.
Currently, we are representing each component in our formal language, KM. We are compiling a dictionary of semantic relations of five types:
Our final step is writing the specifications and axioms for each component. For an event, the axioms include descriptions of the entities that participate in the event, as well as the pre-conditions and post-conditions of it. The axioms are written in our KM language, so they will enable KM to answer a wide range of questions, including ones that require simulation. Follow the link labeled "Browse" (above) to see the specifications and axioms for all our current components.
The Component Library was the cornerstone of our success in the Rapid Knowledge Formation (RKF) project, and played a contributing role in the first phase of Project Halo.
Back to Group Home Page
Created and maintained by Dan Tecuci
Last modified April 19, 2006