Towards a theory of Fa\c{c}ade-X data access: satisfiability of SPARQL basic graph patterns

TL;DR

This paper develops a formal theory for Façade-X data access, focusing on the satisfiability of SPARQL basic graph patterns, and provides algorithms and experiments demonstrating practical feasibility.

Contribution

It formalizes the satisfiability problem for Façade-X, introduces an algorithm for deciding pattern satisfiability, and validates it through extensive experiments.

Findings

Decidable algorithm for basic graph pattern satisfiability on Façade-X data sources.

Practical feasibility demonstrated with real-world queries.

Results support development of efficient data integration systems.

Abstract

Data integration is the primary use case for knowledge graphs. However, integrated data are not typically graphs but come in different formats, for example, CSV, XML, or a relational database. Fa\c{c}ade-X is a recently proposed method for providing direct access to an open-ended set of data formats. The method includes a meta-model that specialises RDF to fit general data structures. This model allows to express SPARQL queries targeting data sources with those structures. Previous work formalised Fa\c{c}ade-X and demonstrated how it can theoretically represent any format expressible with a context-free grammar, as well as the relational model. A reference implementation, SPARQL Anything, demonstrates the feasibility of the approach in practice. It is noteworthy that Fa\c{c}ade-X utilises a fraction of RDF, and, consequently, not all SPARQL queries yield a solution (i.e. are satisfiable) when evaluated over a Fa\c{c}ade-X graph. In this article, we consolidate Fa\c{c}ade-X, and we study the satisfiability of basic graph patterns. The theory is accompanied by an algorithm for deciding the satisfiability of basic graph patterns on Fa\c{c}ade-X data sources. Furthermore, we provide extensive experiments with a proof-of-concept implementation, demonstrating practical feasibility, including with real-world queries. Our results pave the way for studying query execution strategies for Fa\c{c}ade-X data access with SPARQL and supporting developers to build more efficient data integration systems for knowledge graphs.