Never Mind the Semantic Gap: Modular, Lazy and Safe Loading of RDF Data

Eduard Kamburjan, Vidar Norstein Klungre, Martin Giese

Abstract: Any attempt at a tight integration between semantic technologies and object oriented programming will invariably stumble over the gap between the two underlying object models. We illustrate how this semantic gap manifests from the point of view of data retrieval with SPARQL. We present a novel mechanism to load data from RDF knowledge graphs into object-oriented languages that gives static guarantees about the data access and modularly integrates the mapping between the program and the RDF view with the class definition in the program. This allows us to preserve the separation of concerns between the class system of RDF (geared towards domain modeling and data), and that of the program (geared towards typing and code reuse). Loading of RDF can be performed lazily, when required by the program, based on query-futures – subqueries that are only evaluated if and when the data is accessed. We formulate a Liskov principle for the mapping queries to characterize when they respect the subclass relation. Moreover, we provide tool support to detect when the user-provided mapping would cause the loading mechanisms to result in data structures that manifest the semantic gap.