The Role of Constraints in Linked Data

The paper argues that a Linked Data source should publish an application ontology that includes a set of constraints that capture the semantics of the classes and properties used to model the data. Furthermore, if the Linked Data source publishes a mapping between its vocabulary and the vocabulary of a domain ontology, then it has to specify the application ontology constraints so that they are consistent with those of the domain ontology. The main contributions of the paper are methods for constructing the constraints of the application ontology of a Linked Data source, defined as fragments of domain ontologies. The methods assume that the ontologies are written in an expressive family of attributive languages and depend on a procedure to test logical implication, which explores the structure of sets of constraints.