Handling Owl:sameAs via Rewriting
Authors: Boris Motik, Yavor Nenov, Robert Piro, Ian Horrocks
AAAI 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our evaluation shows that our approach can reduce reasoning times on practical data sets by orders of magnitude. Finally, in Section 6 we present a preliminary evaluation of an implementation of our algorithms in the open-source RDFox system. We show that rewriting can reduce the number of materialised triples by a factor of up to 7.8, and can reduce materialisation time by a factor of up to 31.1 on a single thread, with the time saving being largely due to the elimination of duplicate derivations. |
| Researcher Affiliation | Academia | Boris Motik, Yavor Nenov, Robert Piro and Ian Horrocks Department of Computer Science, Oxford University Oxford, United Kingdom forename.lastname@cs.ox.ac.uk |
| Pseudocode | No | The paper describes its algorithm verbally and through an example table but does not present a formal pseudocode block or algorithm figure. It refers to an accompanying technical report for the full algorithm description. |
| Open Source Code | Yes | The implemented system and all test data sets are available online.3 |
| Open Datasets | Yes | Claros has been developed in an international collaboration between IT experts and archaeology and classical art research institutions with the aim of integrating disparate cultural heritage databases.6, DBpedia is a crowd-sourced community effort to extract structured information from Wikipedia and make this information available on the Web.7, Open Cyc is an extensive ontology about general human knowledge.8, Uni Prot is a subset of an extensive knowledge base about protein sequences and functional information.9, Our fifth data set was UOBM (Ma et al. 2006) a synthetic data set that extends the well-known LUBM (Guo, Pan, and Heflin 2005) benchmark. |
| Dataset Splits | No | The paper uses various 'test data sets' for materialization but does not specify any explicit train/validation/test splits or partitioning methodology for these datasets. |
| Hardware Specification | Yes | We conducted our tests on a Dell computer with 128 GB of RAM and two Xeon E5-2643 processors with a total of 8 physical and 16 virtual cores, running 64-bit Fedora release 20, kernel version 3.13.3-201. |
| Software Dependencies | No | The paper mentions 'RDFox' as the implemented system and the operating system '64-bit Fedora release 20, kernel version 3.13.3-201', but it does not provide specific version numbers for ancillary software dependencies such as programming languages or libraries used by RDFox. |
| Experiment Setup | Yes | We have implemented our approach in RDFox, allowing the system to handle owl:same As via rewriting (REW) or the axiomatisation (AX) from Section 3. We then compared the performance of materialisation using these two approaches. In particular, we investigated the scalability of each approach with the number of threads, and we measured the effect that rewriting has on the number of derivations and materialised triples. ... We measured the wall-clock times needed to materialise our test data sets in AX and REW modes on 1, 2, 4, 8, 12, and 16 threads. For each test, we report average wall-clock time over three runs. |