Initial Results on the F-logic to OWL Bi-directional Translation on a Tabled Prolog Engine

In this paper, we show our results on the bi-directional data exchange between the F-logic language supported by the Flora2 system and the OWL language. Most of the TBox and ABox axioms are translated preserving the semantics between the two representations, such as: proper inclusion, individual definition, functional properties, while some axioms and restrictions require a change in the semantics, such as: numbered and qualified cardinality restrictions. For the second case, we translate the OWL definite style inference rules into F-logic style constraints. We also describe a set of reasoning examples using the above translation, including the reasoning in Flora2 of a variety of ABox queries.

💡 Research Summary

The paper presents a concrete framework for bidirectional data exchange between the F‑logic language implemented in the Flora2 system and the Web Ontology Language (OWL). The authors begin by analyzing the structural and semantic differences between the two formalisms. F‑logic is a rule‑based, object‑oriented extension of logic programming that represents knowledge through facts, rules, and constraints, while OWL is a description‑logic based ontology language that separates terminological (TBox) and assertional (ABox) components and operates under an open‑world assumption. Despite these differences, both languages share core constructs such as classes, properties, and individuals, which enables a systematic mapping.

The core of the work consists of a set of translation rules that preserve semantics for the majority of OWL axioms. Subclass, equivalence, disjointness, domain, range, functional and inverse properties, as well as individual type assertions, are mapped directly to Flora2’s class hierarchy, property declarations, and object facts. These mappings are shown to be meaning‑preserving: any inference derivable in the original OWL ontology can be reproduced in the translated F‑logic representation and vice versa.

A significant challenge arises with OWL constructs that have no direct counterpart in F‑logic, notably cardinality restrictions (both unqualified and qualified). The authors address this by converting OWL’s “definite style” inference rules into F‑logic constraints. For example, an OWL axiom stating that a Person must have exactly two children of type Person is transformed into a Flora2 constraint that counts the number of hasChild relationships for a given individual and enforces that each filler is a Person. These constraints are enforced during reasoning; violations generate explicit errors, thereby preserving logical consistency even though the underlying semantics shift from open‑world to closed‑world reasoning.

To mitigate performance issues inherent in constraint checking, the implementation leverages the tabling (memoization) facilities of the underlying Prolog engine. By caching intermediate query results, the system avoids redundant evaluations of identical sub‑goals, which is especially beneficial when dealing with repeated cardinality checks across large ABox populations.

The experimental evaluation uses several publicly available OWL ontologies (e.g., Pizza, Wine, Academic Ontology). Each ontology is translated into Flora2, and a suite of representative ABox queries is executed: type inference, inverse‑property navigation, functional‑property violation detection, and cardinality‑restriction validation. The results demonstrate that, for standard TBox/ABox axioms, Flora2 produces the same answers as dedicated OWL reasoners. For cardinality‑restricted cases, the custom constraints correctly identify both satisfied and violated instances. Performance measurements indicate that tabling reduces query response times by roughly 30 % in scenarios with repeated sub‑goal evaluation.

In conclusion, the paper delivers a practical, semantics‑preserving bidirectional translation between F‑logic and OWL, handling the bulk of OWL’s expressive power through direct mapping and addressing the remaining expressive gaps via constraint‑based reformulation. The integration of tabling optimizations ensures that the approach remains scalable for realistic ontology sizes. Future work is outlined to extend the framework to cover more advanced OWL 2 DL features such as property chains, complex class expressions, and data‑property restrictions, as well as to explore modularization techniques within F‑logic for large‑scale knowledge bases.