Assimilation: A Strategy for Implementing Self-Reorganizing Knowledge Bases

Book One

Volume

Proceedings of the AAAI Conference on Artificial Intelligence, 6

Knowledge Representation

Assimilation is a process by which a knowledge base restructures itself to improve the organization of and access to information in the base. This paper presents a strategy for implementing assimilation in propositional knowledge bases which distinguish between the axioms of the system’s knowledge (called the context) and the derived consequences of those axioms (called the belief space). The strategy in question takes advantage of housekeeping phases in which the system discards accumulated clutter to discover useful patterns of access on the basis of which the context can be reorganized. Unused axioms are replaced by their more useful consequences; derivable generalizations that shorten common inference paths are added to the belief space.