Data and Knowledge Group

― Knowledge Representation and Reasoning

RDFox Tests

Welcome to the RDFox evaluation page. This page is in support of the submission Parallel Materialisation of Datalog Programs in Centralised, Main-Memory RDF Systems to the AAAI 2014 conference. A full version of the paper can be found here. This page contains downloads for the RDF stores RDFox and DBRDF, the datasets used in our evaluation, and the complete test results of our evaluation. Innovation. After the initial development at University of Oxford, RDFox is now ready for production and is available commercially from Oxford Semantic Technologies, a spin-out of the University backed by Samsung Ventures and Oxford Sciences

Systems

Our evaluation included the RDF stores RDFox, DBRDF (with its four versions PostgresSQL-VP, MonetDB-VP, PostgresSQL-TT, and MonetDB-TT), and OWLIM-Lite. For a detailed description of these systems, please refer to Section 5 of the paper and the references therein. Please find details for acquiring these systems below.

RDFox

The experiments were conducted with the following version of RDFox.zip, distributed under the following academic licence. The zip file contains libraries and executables which are precompiled for 64-bit Red Hat Enterprise Linux Server 6.3, kernel version 2.6.32. Additionally, the zip file contains the source code which can be viewed and compiled using the Eclipse IDE. Usage instructions can be found here.

DBRDF

DBRDF can be downloaded as an Eclipse project from here and as a library from here. In our evaluation, we used the following shell scripts for importing RDF data, for reasoning with MonetDB, and reasoning with PostgreSQL.

OWLim

In our evaluation, we used OWLim Lite v5.3. Usage instructions can be found here.

Test Datasets

We conducted our evaluation with the synthetic datasets LUBM and UOBM, and the real-world datasets DBPedia and Claros. The files for each dataset can be found here and are organised in folders as follows:

  • ttl contains html files which redirect to ziped RDF graphs for the dataset (i.e. different ABoxes);
  • owl contains the original and derived ontologies (i.e. the TBox);
  • dlog contains rulesets derived from the ontologies in owl in format suitable for RDFox;
  • pie contains rulesets derived from the ontologies in owl in format suitable for OWLim;
  • sql contains sql scripts generated and used by DBRDF during reasoning.

For more details on the test data refer to Section 5 of the paper.

Ruleset extensions

As pointed out in Section 5 of the paper, we extended some of the ontologies with manually created rules. These rules can be found at the following links for the rulesets LUBM, DBPedia and Claros.

Test Results

We have evaluated the scalability of parallel reasoning with RDFox on the ARCUS system of the Oxford Supercomputing Centre. In support of Section 5 in the paper, we provide the test results for the scalability of RDFox for each of the three runs run1 (pdf, csv), run2 (pdf, csv), and run3 (pdf, csv), together with average results (pdf, csv), and a summary of the results.

We also compared the sequential version of RDFox with competitor systems on our in-house Linux server KRR. In support of Section 5 in the paper, we provide our full evaluation results for the stores RDFox (pdf, csv), PostgresSQL-VP (pdf, csv), MonetDB-VP (pdf, csv), PostgresSQL-TT (pdf, csv), MonetDB-TT (pdf, csv), OWLIM (pdf, csv).

Using RDFox

Researchers

Boris Motik, Yavor Nenov, Robert Piro, Ian Horrocks

Key Publications