Using Expert Knowledge in Database-Oriented Problem Solving

Database-oriented problem solving often involves the processing of deduction rules which may be recursive in relational database systems. In this kind of problem solving, expert knowledge plays an important role in the guidance of correct and efficient processing. This paper presents a modularized relational planner RELPLAN, which develops a knowledgedirected in ference and planning mechanism for efficient processing of deduction rules in relational DB systems. Introduction Relational database technology provides us with a powerTo effectively augment expert knowledge in relational ful tool for in formation processing. The conventional use DB system. a relational problem solving planner of relational DB systems is aimed at the management and RELPLAN has been built. Similar to many expert sysretrieval of stored data. With the emerging research on tems, expert knowledge is encoded in RELPLAN in the expert database systems, expert system technologies are form of rules and incorporated with queries in deductive being merged into relational database systems and the compilation to answer queries and solve problems. In our application domains of relational DB technology are design, the modularization of a rule system and the combeing expanded to those that require knowledge-guided pilation technique are emphasized. A modularized rule processing of both stored and derived data. system is built on top ofaconventional relational DB system and RELPLAN uses these rules to transform user's This paper studies the application ofexpert knowledge in deductive queries into non-deductive query programs, DB-oriented problem solving. Problem solving is the functioning as a deductive front-end of the relational DB process of developing a structure of (in the simpiest case, system. a sequence 00 actions to achieve a goal. Databaseoriented problem solving is the problem solving involvComplex DB-oriented problem solving requiresplanning ing targe databases, in our discussion, large relational technique, which develops a structure of query plans databases. As in many expert systems, DB-oriented (programs) for a problem before actually solving it by problem solving is featured with deductive process. Our DB operations. The planning technique implemented in discussion is more concentrated on the deductive process our project is the means-ends analysis technique which which may involve recursive rules. develops hierarchical plans for complex queries based on the modification of the original deductive module. The To efficiently implement such problem solving process in planning process is divided into two phases: the selection relational databases, two issues should be addressed. The of a planning strategy and the generation of the actual first one is the transformation of recursion into iteration plan. The selection of planning strategy is based on the in relational databases. Two recent research papers query provided by database user and the information &Hens 846 and &Ullm 856 deal with this problem from stored in the database. two differeni angles. &Ullm 856 develops query evaluatien techniques based on "capture rules" an a graph This paper first illustrates the architecture of the relationrepresenting clauses and predicates, while [Hens 84] at planner RELPLAN, then discusses the compilation of presents us algorithms which compile queries involving non-recursive and recursive queries using expen knowlrecursive rules into iterative programs. The second issue edge. A two-phase planning mechanism using plan is how to use expert knowledge and planning techniques modules is introduced and the efficiency of knowledgeto guide efficient processing of recursion or iteration in directed inference and planning in database-oriented relational databases. This is the topic of this paper. problem solving is also demonstrated in the paper.