Knowledge Engineering under Time and Uncertainty
From Distributed Information and Intelligence Analysis Group
We work continually on developing new knowledge engineering models, frameworks, and tools. We are especially interested in reasoning under uncertainty. This can be uncertainty about the truth of a proposition or when an event may occur. The primary framework we employ is that of Bayesian Knowledge Bases (BKBs), a modeling framework that subsumes Bayesian Networks. There are a few main areas we are focusing on in this work:
- Temporal Knowledge - BKBs have been extended to support temporal knowledge in our Temporal Bayesian Knowledge Bases (TBKBs) and work continues to enrich the representational framework.
- Knowledge Fusion - An especially difficult problem in probabilistic modeling is the fusion, or aggregration, of information. In 2008 we introduced our Bayesian Knowledge Fusion algorithm which has been shown to be probabilistically valid.
- Verification and Validation - A key component of any knowledge representation framework is the ability to perform verification and validation. Our work has focused on how to ensure that reasoning over a knowledge base produces the expected results. In the event that it does not, the system should identify why this is happening and where additional knowledge could be added to remedy the situation.
