Structural validation of Enhanced Entity-Relationship models using description logic reasoners

Abstract

The Enhanced Entity-Relationship (EER) language is widely used in the creation of conceptual database models. The validation of these models is critical as validity erros can be passed to the next phases of the project and negatively influence the outcome. In large and complex models, validation becomes a difficult task because the interaction between the elements used can produce inconsistencies and unintended implicit consequences. Hence, it is essential to offer automatic assistance. Description Logics (DLs) are a set of languages used for knowledge representation. They admit decidable and automated reasoning tasks, such as the identification of implicit logical consequences. Because of those characteristics, DLs have been considered a promising alternative to represent and reason on conceptual models. This work aims to support the validation of conceptual database models by identifying syntactic and semantic inconsistencies in EER models using DL reasoners. To the best of our knowledge, few work use Description Logics to represent and reason on EER models. Also, these work do not cover aspects such as the interaction between model constraints and the related structural consequences. Our work stands out for taking into account the consequences of constraints such as cardinality, participation, relationship type degree, inheritance, cyclic paths, and valid attribute types, as well as the consequences of the interactions between these constraints on the same model. With the support of Protégé, we built a Knowledge Base(KB) in OWL DL by formalizing the EER syntax. Next, we added the semantic validity rules related to the constraints mentioned. Although we tried to represent most of the rules by using axioms, we also made use of Semantic Web Rule Language (SWRL) rules in cases in which DL expressivity was not sufficient. Finally, we manually converted the KB to ALCROIQ language. As proof of concept, we successfully validated case studies by using DL reasoners.