Intensional Query Processing in Deductive Database Systems.

Abstract

This dissertation addresses the problem of deriving a set of non-ground first-order logic formulas (intensional answers), as an answer set to a given query, rather than a set of facts (extensional answers), in deductive database (DDB) systems based on non-recursive Horn clauses. A strategy in previous work in this area is to use resolution to derive intensional answers. It leaves however, several important problems. Some of them are: no specific resolution strategy is given; no specific methodologies to formalize the meaningful intensional answers are given; no solution is given to handle large facts in extensional databases (EDB); and no strategy is given to avoid deriving meaningless intensional answers. As a solution, a three-stage formalization process (pre-resolution, resolution, and post-resolution) for the derivation of meaningful intensional answers is proposed which can solve all of the problems mentioned above. A specific resolution strategy called SLD-RC resolution is proposed, which can derive a set of meaningful intensional answers. The notions of relevant literals and relevant clauses are introduced to avoid deriving meaningless intensional answers. The soundness and the completeness of SLD-RC resolution for intensional query processing are proved. An algorithm for the three-stage formalization process is presented and the correctness of the algorithm is proved. Furthermore, it is shown that there are two relationships between intensional answers and extensional answers. In a syntactic relationship, intensional answers are sufficient conditions to derive extensional answers. In a semantic relationship, intensional answers are sufficient and necessary conditions to derive extensional answers. Based on these relationships, the notions of the global and local completeness of an intensional database (IDB) are defined. It is proved that all incomplete IDBs can be transformed into globally complete IDBs, in which all extensional answers can be generated by evaluating intensional answers against an EDB. We claim that the intensional query processing provide a new methodology for query processing in DDBs and thus, extending the categories of queries, will greatly increase our insight into the nature of DDBs.