Processing deductions in an active fashion

Abstract

Active and deductive rules in databases and procedural attachment in knowledge bases are used as mechanisms of computation of derived attributes. These research domains have become very closely related. The support of derivations in current active databases suffers from many semantic and technical problems. Active rules in their ECA (Event Condition Action) form are event oriented whereas derivations are naturally data oriented. Deductive rules are well adapted to relational databases, but their integration within the object oriented model does not always have successful results. Procedural attachment may be considered as both a deductive and active process, but it uses an imperative language to derive attributes. In this paper, we try to focus on the use of active rules to derive attributes in an object oriented model. We give our approach for the support of deductions in a declarative assertion language and their processing in an active way. We propose a declarative derivation approach integrated within an object oriented model. IntoductionAn active database is a database which can automatically carry out some predefined actions in response to some specific events when the corresponding conditions are satisfied [15] [20]. Active databases appeared when researchers in the database field recognized the use of the integration of production rules in relational DBMSs [27]. Most active database models use an ECA approach in which when the event occurs the condition is tested, if it is satisfied, the action is executed. These rules have their roots in artificial intelligence with the languages of expert systems such as OPS5[lo]. They have been justified as a suitable framework for the specification of queries, views, integrity constraints, and triggers in databases [22]. In contrast, these rules are not the best choice to support derivations, which are naturally data oriented. The values of derived attributes are expressed as functions of values of other attributes. The computation tools of these attributes are invoked when the attributes they depend on are modified. The use of active rules for supporting derivations does not take advantage of this dependency relationship between derived attributes and the other attributes. Meanwhile, using another mechanism, we can take advantage of this dependency, for better performance at the design and execution levels, and to be able to analyse an application behaviour by determining the consequences of each modification. Deductive databases constitute a declarative natural manner to enrich relational databases by intelligent processing (deduction). The origin of deductive databases goes back to work on theorem proving [25][31]. They were introduced as extensions of relational databases by logic. Deduced facts are the logical consequences of the previous facts using the specified deduction rules. The integration of these rules and the object oriented models creates impedance mismatch problems [16]. In fact, the object oriented concepts are difficult t o introduce in the deductiv...