Making Query Processing Goal-Directed in Disjunctive Logic Databases

Disjunctive logic programs have recently attracted much attention. While a large body of research has been recently acquired on the semantic issues, much work is still be needed to obtain efficient query processing procedures in disjunctive logic programs. In the case of definite programs, many efficient query processing strategies both topdown and bottom-up are well known. Their efficiency mainly comes from both avoiding redundancy and goal-direcredness. Unlike definite programs, disjunctions in the head of a rule, however, make the query processing much complicated. We propose a goal-directed query processing procedure for disjunctive logic databases. To do that, we utilize the notion of “relevancy” which was proposed in a theorem prover called SATCHMORE [LRW93]. Though useful for the satisfiability checking of a given set of clauses, the notion of relevancy is not sufficient as it is for the query processing in disjunctive databases, where it is essential in some applications for the answer to a given query to be distinguished between “true”, “false”, or “unknown (possibly true)”. We first give a goal-directed query processing procedure based on the relevancy, but suitably modified for the above purpose, and show that our procedure is sound and complete for a class of positive disjunctive databases in the minimal model semantics. We then show that a modified version of our procedure effectively works as a correct query processing method for the possible model semantics [S193], a recently proposed alternative semantics for disjunctive databases. We also give some experimental results of our query processing procedure. which show that a significant performance improvement has been brought about compared with naive bottom-up computation.