Position Paper: Ontological Logic Programming

Murat Sensoy, Geeth de Mel, Wamberto Weber Vasconcelos, Timothy J. NormanSSN 2010

Introduction

In knowledge representation, a fundamental tension exists between the formal decidability of Description Logics (DL) and the expressive flexibility of Logic Programming (LP). While the Web Ontology Language (OWL) excels at defining hierarchical domain knowledge, it often lacks the procedural power required for complex data structures and "negation as failure." Conversely, Logic Programming languages like Prolog are highly expressive but do not natively interpret the rich semantics of modern ontologies. This paper introduces Ontological Logic Programming (OLP), a framework designed to bridge this gap by integrating OWL-DL reasoning directly into a Prolog-based execution environment.

The OLP Framework

The core of OLP is a meta-interpreter that functions as a bridge between a Prolog engine (tuProlog) and a Description Logic reasoner (Pellet). The system architecture is organized into a layered stack that allows for the delegation of ontological queries during runtime.

The OLP system architecture stack, showing the integration of Prolog and DL reasoners.

When the OLP interpreter encounters a predicate with a namespace prefix (e.g., vin:hasFlavor), it identifies it as an "ontological predicate." Instead of searching the local Prolog database, the interpreter queries the DL reasoner. The reasoner utilizes imported OWL ontologies and SWRL rules to infer facts, which are then passed back to the Prolog engine to continue execution. This native integration allows developers to use ontological terms directly in their code without manual translation, leveraging the ontology to prune search spaces by understanding domain-specific dependencies.

Performance Evaluation

The OLP framework was evaluated using a "sensor-task matchmaking" scenario based on the ISTAR (Intelligence, Surveillance, Target Acquisition, and Reconnaissance) ontology. This problem involves finding optimal combinations of platforms and sensors to meet specific mission requirements. The evaluation used a synthetic dataset of 908 tasks to compare OLP against a traditional exhaustive search algorithm.

Performance comparison between OLP and exhaustive search as configuration complexity increases.

The results demonstrate that OLP significantly outperforms exhaustive search. As the complexity of configurations (Maximum DC Size) increases, the exhaustive search time grows exponentially. In contrast, the OLP approach maintains a nearly linear growth in running time. By using the ontology to rule out invalid combinations early in the reasoning process, OLP transforms a computationally expensive search into a highly efficient task.

Conclusion and Future Work

Ontological Logic Programming provides a practical solution for applications requiring both deep domain knowledge and complex logical constraints. By combining the strengths of the Semantic Web and Logic Programming, OLP enables efficient resource allocation and matchmaking. Future work will focus on enhancing the interpreter's explanation facilities, providing users with detailed justifications for the system's conclusions and further exploring performance optimizations for large-scale ontologies.

Summary updated

BibTeX
@inproceedings{sensoy2010position,
  author = {Murat Sensoy and Geeth de Mel and Wamberto Weber Vasconcelos and Timothy J. Norman},
  title = {{Position Paper: Ontological Logic Programming}},
  booktitle = {SSN},
  year = {2010},
}
Figure