Object space partitioning in a DL-like database and knowledge base management system

Abstract

The p-type data model was designed first to answer database needs. Some of its features were and still are quite unusual for a DBMS and, by some aspects, make it nearer Description Logic (DL) systems than classical DBMS. Views play a central role in the model. They are defined in a hierarchical manner, with constraints on role (attribute) types as in DLs, and instance classification (view recognition) is a basic mechanism in p-type implementation by the Osiris system. In this paper we recall the main characteristics of p-types and their semantics as a DL system. We insist on the modelling of unknown values, whose treatment leads to a three-value instance classification system. We develop database specific aspects and particularly the partitioning of the object space and its use for the management of data.