- The paper proposes a direct mapping of relational databases to RDF and OWL using Datalog rules, aiming for key properties like information and query preservation.
- It details essential properties such as information preservation (reconstruct data) and query preservation (SQL-to-SPARQL) and explores challenges in achieving monotonicity and semantics preservation, especially with foreign keys.
- The authors suggest that while standard OWL struggles with relational constraints, extensions or non-monotone mappings might achieve better semantic fidelity, paving the way for future work on datatype and scale considerations.
Direct Mapping of Relational Databases to RDF and OWL
The paper "On Directly Mapping Relational Databases to RDF and OWL (Extended Version)" examines the translation of relational databases into RDF graphs with OWL vocabulary, a key issue for the advancement of the Semantic Web. This research focuses on deploying a set of Datalog rules to map relational databases while addressing certain foundational data properties and desirable semantic integrity.
Fundamental and Desirable Properties
The authors introduce essential properties for effective mapping: information preservation and query preservation. Information preservation ensures that original relational data can be reconstructed from its RDF representation, a critical aspect of data integration workflows. Query preservation guarantees the convertibility of relational queries into equivalent RDF queries, facilitating workloads that demand SQL-to-SPARQL translations.
Moreover, the paper discusses two desirable properties: monotonicity and semantics preservation. Monotonicity allows incremental updates without reevaluating the entire database, thereby optimizing performance. Semantics preservation ensures the mapping reflects database integrity constraints, encoding violations in the output RDF graph as inconsistencies.
Direct Mapping Approach
The suggested solution leverages Datalog rules for a direct mapping that is monotone, information-preserving, and query-preserving, even when handling relational databases with null values. The proposed mapping corrects monotonicity issues when striving for semantic preservation by employing additional rules to detect and reflect database inconsistencies (such as primary key violations) within the RDF results.
Handling Primary Keys and Foreign Keys
Semantic preservation demands tackling both primary key and foreign key constraints. For primary keys, the paper extends the direct mapping to identify key violations, thereby ensuring consistency between the relational and RDF representations. However, the integration of foreign keys proves more challenging; it is demonstrated that no monotone mapping can achieve semantic preservation due to inherent OWL characteristics like Open World Assumption (OWA).
To overcome this, the authors present a non-monotone mapping that introduces artificial triples, making RDF graphs semantically representative of relational constraints. Additionally, they explore extending OWL with epistemic operators to maintain monotonicity alongside semantic preservation, hence aligning closed-world relational assumptions with open-world OWL semantics.
Implications and Future Directions
The findings underscore that while the current OWL semantics don't naturally support integrity constraints from relational models, theoretical extensions in OWL incorporating epistemic logic might provide a path forward for full semantic fidelity between relational databases and their semantic web representations.
Future work should address extending data mappings to include datatype considerations, bag semantics, and evaluating larger-scale implementations. Integrating RDF/OWL transformations that maintain both information fidelity and semantic correctness is critical for broader adoption in data interoperability scenarios.
The paper serves as a foundational paper highlighting essential problems and potential solutions in the field of database-to-semantic web mappings, paving the way for further research on sophisticated data translation methodologies.