A query and inference system based upon knowledge indexed deductive search
Kuk Hie Lee · 1990
A query inference system, called KIDS, has been developed. Its design involves the incorporation of well-developed database techniques (i.e., indexing and query optimization) into the technology of artificial intelligence (i.e., representation and inference), which would be suitable from the point of view of both expressiveness and efficiency. This dissertation describes some salient considerations that led to the design of KIDS. This research addresses the issue of how to represent complex-structured real-world information. The KIDS representation formalism exploits named property lists in a manner that does not compromise the well-defined syntax and semantics of first-order logic. It allows a more natural and flexible way of expressing information, beyond that of the relational query languages and the predicate logic syntax. Furthermore, the formalism is conducive to designing a sophisticated indexing scheme which can provide fast access to relevant information. A novel feature of this research is the use of the indexing scheme for controlling the deduction. When there exists more than one alternative to pursue (e.g., the existence of more than one literal in a disjunction or rule and the existence of several relevant rules), the indexing scheme provides control information, which guides the system in its choice of which alternative to pursue. Advantages of the index-guided approach are improved system performance and better system control of inferences.