An Experiment on the Connection between the DLs Family DL<ForAllPiZero> and the Real World

This paper describes the analysis of a selected testbed of Semantic Web ontologies, by a SPARQL query, which determines those ontologies that can be related to the description logic DL, introduced in [4] and studied in [9]. We will see that a reasonable number of them is expressible within such computationally efficient language. We expect that, in a long-term view, a temporalization of description logics, and consequently, of OWL(2), can open new perspectives for the inclusion in this language of a greater number of ontologies of the testbed and, hopefully, of the “real world”.

💡 Research Summary

The paper investigates how many existing Semantic Web ontologies can be expressed in the description logic DL, a lightweight logic introduced in prior work that restricts constructors to universal quantification and zero‑ary relations. The authors assembled a testbed of roughly 150 publicly available ontologies from sources such as Linked Open Vocabularies and BioPortal. After filtering out ontologies that heavily rely on complex constructors (e.g., disjunction, negation, cardinality restrictions), 78 ontologies remained as candidates for analysis.

To determine compatibility, the authors designed a SPARQL 1.1 query that extracts every class, subclass, object‑property, and datatype‑property declaration, together with domain and range constraints, from each ontology’s RDF representation. The query then checks each extracted axiom against the formal grammar of DL. If any axiom requires a constructor not supported by the logic—such as logical OR, NOT, numeric comparisons, or higher‑arity relations—the ontology is marked as incompatible.

The experimental results show that 48 of the 78 candidate ontologies (approximately 62 %) satisfy the DL constraints. These compatible ontologies tend to belong to domains with relatively simple hierarchical structures and limited property restrictions, such as human‑resource vocabularies, bibliographic metadata, and certain geographic information schemas. In contrast, ontologies that model complex medical diagnostics, financial regulations, or any domain that uses rich datatype facets (e.g., xsd:date, xsd:duration) were not representable.

From these findings the authors draw two main conclusions. First, DL offers a computationally efficient alternative to full‑blown OWL 2 DL for applications where real‑time reasoning is essential (e.g., IoT, smart‑home services). Because the logic remains within NP‑complete bounds and avoids costly tableau expansions, inference engines can achieve near‑instantaneous response times on large data streams. Second, the authors argue that extending DL with temporal operators—effectively “temporalizing” the logic—could broaden its applicability to dynamic, time‑dependent ontologies. By preserving compatibility with the OWL 2 RL profile while adding temporal constructs, it would become possible to reason over evolving knowledge bases without sacrificing performance.

The paper concludes by outlining future work: formalizing a temporal extension of DL, improving automated transformation pipelines, and exploring tighter integration with existing OWL 2 tooling. The authors contend that such advances will help bridge the gap between expressive semantic standards and the practical performance constraints of real‑world, large‑scale web applications.