Knowledge Acquisition for Level II/III Fusion

"This proposal focuses on the development of algorithms and methodologies for the acquisition of probabilistic relational knowledge from subject-matter experts with little or no training in knowledge representation and probability theory.The knowledge representation effort today follows two major threads of work. The first one involves the creation of tools for efficiently maintaining and manipulating large-scale Ontologies and Knowledge Bases. The second focuses on uncertainty handling inthe knowledge representation. This latter thread involves the assertion of the Bayesian Networks paradigm. Traditionally, the Ontology and Knowledge Bases representation languages allow the encoding of complex relational knowledge but do not support anefficient and sound way of handling uncertainty issues. Conversely, Bayesian Networks do handle uncertainty in a proper way, but they lack of the powerful expressiveness of relational models.Recent advances in the Bayesian Networks paradigm, namely Probabilistic Relational Models, have built a bridge between the integration of the sound uncertainty handling techniques of Bayesian Networks into the Ontology and Knowledge Bases developmentsystems. This integration will have a synergic effect, exploiting the complementary expressive power of both the systems. With this proposal we intend to analyze and solve the issues related to this integration effort. The Integrated Knowledge Development System will all