Description Logic Reasoning under Uncertainty

From Distributed Information and Intelligence Analysis Group

Ontologies are powerful knowledge representation tools which can be used to formally describe intuitive concepts, individual entities, and the attributes and properties which define them. They are a way for machines to represent and reason about conceptual knowledge in a way intuitive to humans. Unfortunately, ontologies' underlying theory, Description Logic is limited in that it cannot reason about uncertainty. To date, work on uncertainty frameworks for Description Logic has not intuitively captured a useful notion of uncertainty, for reasons including weaknesses in underlying uncertainty models and assumption conflicts with semantic networks.

We have developed, and are currently refining, a knowledge theory which is a complete and intuitive synthesis of Description Logic and Bayesian Knowledge Bases. This synthesis represents knowledge as “if-then” conditional probability rules between description logic assertions. It has already been shown to overcome the weaknesses and conflicts of past attempts, and promises to increase the power, flexibility, and ease of use of existing knowledge modeling tools when implemented. We are also currently developing novel reasoning and validation capabilities based on Bayesian Knowledge Fusion.