A hybrid approach to developing ontology-driven applications : a case study in describing radio spectrum usage policies

Abstract

As such, applications need a simple and stable API that abstracts much of the structure, but keeps the strengths of OWL in reasoning or clarity. To meet this need, we utilize a hybrid approach to integrating an ontology into domain-specific applications, specifically in the domain of computable policies. A hybrid approach is characterized by choosing what to model in domain-specific classes, while delegating the rest to the underlying knowledge graph. Our approach uses aspects of domain-driven design and a backing domain-specific language to capture the essence of a domain with links to the ontology to preserve the strengths of OWL. To analyze this approach, we describe an implementation of a library that models radio spectrum usage policies in the Dynamic Spectrum Access Policy Framework and discuss it's strengths and weaknesses.

The Semantic Web initiative pushes for use of open specifications (e.g. Resource Description Framework (RDF) and Web Ontology Language (OWL)) to describe data and support automated inference from machine-readable logical rules in Ontology-Driven Applications. In such applications, the domain-specific data is often directly parsed from the graph-based data model. However, when the code and the ontology are tightly linked, the flexible nature of ontology development makes code maintenance difficult in the long term. For example, the structure of an ontology often changes to add more semantics or to enable automated reasoners to apply complex business logic to the data (e.g. reasoning with OWL and HermiT). In many cases, the underlying domain data is left unchanged. Regardless, these changes require developers to update their code to accommodate the new structure. When multiple applications directly rely on the graph, changes propagate across all of them, further slowing down application development and increasing load on the developer. Additionally, with OWL-based ontologies, extensions such as meta-modelling introduce further difficulty when extracting information from an ontology, as it needs complex graph traversal code. This encourages the growth of technical debt over time.