Reporting on RDF* (and SPARQL*) engines' behaviours…
In 2014, Hartig and Thompson proposed RDF*: a syntax to express statements of statements for RDF data without involving tedious reification methods. Furthermore, following this RDF extension, they also suggested an extension based on SPARQL for querying these new datasets. Recently, Hartig shared an extension for the SPARQL UPDATE fragment.
Driven by the propositions from Hartig and Thompson, multiple efforts have been realized by the Semantic Web community so far:
- First of all, Hartig developed a set of tools based on top of Jena.
- Blazegraph.
- AnzoGraph.
- In December 2019, Stardog.
- GraphDB.
- In May 2020, RDF4J (former Sesame) announced that version 3.2 provides an experimental support.
- Two weeks later in May 2020, Apache Jena released its version 3.15.0 which now supports RDF*.
As presented above, several popular RDF/SPARQL solutions provide a support for the aforementioned extensions. However, as there is yet no standard approaches, engines have a high degree of freedom. Based on that observation, we wanted to review the current solutions developed and shed light on their possible differences. In a nutshell, our goals were: first to observe the existing landscape, developping methods and protocols to compare the solutions together; and second to discuss the results and share our findings with the hope they would help the community.
As a first study, we designed a simple experimental protocol in order to investigate the internal representations that RDF* compliant engines have.
This project is shared under the terms of the Apache License v2.0 [here].
Fabrizio Orlandi
Damien Graux
Declan O'Sullivan
Affiliated to the ADAPT Centre from the Trinity College Dublin in Ireland.
Research partially conducted as part of an EDGE Marie Sklodowska-Curie fellowship, grant agreement No. 713567.