Uniqueness constraints in object-relational databases and description logics

Abstract

We address problems that arise in many areas of information technology. In particular, our work considers how to effectively represent semantic constraints commonly used in object-relational database systems and then develops efficient algorithms to reason about such constraints. We expand data dependencies that are commonly used by relational models and combine them with constraints arising in object-relational models to form a theory that allows one to reason about both kinds of constraints. We then present procedures that can efficiently reason in such a theory. The procedures can be used to help solve problems relating to both object-relational and relational databases. A fundamental reason that we are able to derive such procedures relates to the variety of uniqueness constraints incorporated into the theory, which strictly generalize the notions of keys and functional dependencies that are inherent parts of relational and object-relational models.

We investigate the interaction between such constraints and other data dependencies, including inheritance, typing and equational constraints. In addition, the problem is explored in the context of description logics (DLs), which are a family of knowledge representation schemas that have found myriad application in information systems technology. From this perspective, we introduce a new concept constructor for capturing uniqueness constraints within the DL framework, show how various DL dialects that include this new constructor can be used to address problem in information technology and present efficient decision procedures for subsumption checking in these dialects.

Among other contributions, we further analyze our procedures with respect to their generality, incremental abilities, and other characteristics.