[Paper Review] Reverse Engineering Ontology to Conceptual Data Models

Haya El-Ghalayini, Mohammed Odeh+2 more|ArXiv.org|Dec 8, 2004

Semantic Web and Ontologies16 references10 citations

TL;DR

This paper proposes a method to reverse-engineer DAML+OIL ontologies into domain conceptual data models using semantic mappings derived from the BWW model. By applying transformation rules to the TAMBIS ontology, the study demonstrates that a valid conceptual model can be automatically generated, though domain expert validation is required before evolving it into a global model.

ABSTRACT

Ontologies facilitate the integration of heterogeneous data sources by resolving semantic heterogeneity between them. This research aims to study the possibility of generating a domain conceptual model from a given ontology with the vision to grow this generated conceptual data model into a global conceptual model integrating a number of existing data and information sources. Based on ontologically derived semantics of the BWW model, rules are identified that map elements of the ontology language (DAML+OIL) to domain conceptual model elements. This mapping is demonstrated using TAMBIS ontology. A significant corollary of this study is that it is possible to generate a domain conceptual model from a given ontology subject to validation that needs to be performed by the domain specialist before evolving this model into a global conceptual model.

Motivation & Objective

• To address semantic heterogeneity in heterogeneous data sources by leveraging ontologies for integration.
• To investigate whether a domain conceptual data model can be automatically generated from an existing ontology.
• To establish a formal mapping between ontology elements (in DAML+OIL) and conceptual model constructs.
• To validate the feasibility of evolving the generated model into a global conceptual model through expert review.
• To demonstrate the approach using a real-world ontology, TAMBIS, as a case study.

Proposed method

• Define a semantic mapping framework based on the BWW model to relate DAML+OIL ontology constructs to conceptual data model elements.
• Identify transformation rules that map ontology components such as classes, properties, and axioms to conceptual model entities, relationships, and constraints.
• Apply the mapping rules to the TAMBIS ontology to generate a corresponding conceptual data model.
• Use formal semantics from DAML+OIL to ensure consistency and correctness during the reverse engineering process.
• Validate the generated model through expert review to ensure semantic fidelity before global model integration.
• Demonstrate the approach with a case study involving the TAMBIS ontology and its transformation into a conceptual model.

Experimental results

Research questions

• RQ1 Can a domain conceptual data model be automatically generated from an existing DAML+OIL ontology?
• RQ2 What semantic mappings are necessary to ensure fidelity between ontology constructs and conceptual model elements?
• RQ3 How can the reverse-engineered conceptual model be validated for accuracy and usability?
• RQ4 What role does domain expert validation play in evolving a local model into a global conceptual model?
• RQ5 To what extent can the BWW model serve as a foundation for ontology-to-conceptual model transformation?

Key findings

• A formal mapping between DAML+OIL ontology elements and conceptual data model components was successfully established using the BWW model.
• The transformation process successfully generated a domain conceptual model from the TAMBIS ontology.
• The generated model required validation by a domain specialist before it could be considered suitable for integration into a global model.
• The study confirms that reverse engineering from ontology to conceptual model is feasible under defined semantic mappings.
• The approach demonstrates potential for scalable data integration by enabling automated model generation from semantic ontologies.
• The results indicate that while automation is possible, human expertise remains essential for ensuring model correctness and semantic alignment.