Large deductive databases with constraints

This thesis addresses the problem of building a knowledge base management system, or more precisely, a deductive database management system (DDBMS). In a deductive database there are three kinds of information: facts, logical deduction rules and constraints. A DDBMS should provide the functionality of a modern relational DBMS, and in addition, it should handle rules as well as arbitrary integrity constraints. First, we introduce the mathematical background of deductive databases. This background is based on first order logic. Subsequently, we outline the basic decisions we have taken to build a DDBMS. To realize such a system, a new algorithm has been developed to process rules, constraints and queries involving quantifiers. As a next step we turn to the problem checking the consistency (semantic integrity) of a deductive database in the face of update transactions involving one or more update operations. We present a new method to test whether the constraints are still satisfied after an update. In contrast to the methods presented in the literature, our method can deal with transactions which involve the addition and/or deletion of an arbitrary number of facts and even rules. Moreover any kind of semantic integrity constraint may be handled. Afterwards, the notion of deductive database is extended so that temporal aspects of the knowledge may be represented as well. In particular, we investigate the behavior of constraints involving time operators. In addition to developing efficient algorithms and implementation techniques to manage a deductive database, it is possible to improve the efficiency of the processing of queries and updates on a high level of abstraction, namely by transforming knowledge from its given form into equivalent simpler forms. We investigate, in a more general way than other studies, what kinds of knowledge transformations are possible without changing the semantics of the knowledge base. An overview of the implementation of a DDBMS is given. This implemen¬ tation realizes the concepts presented.