Datalog Rewritability of Disjunctive Datalog Programs and its Applications to Ontology Reasoning

Authors: Mark Kaminski, Yavor Nenov, Bernardo Cuenca Grau

AAAI 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our empirical results suggest that many non-Horn ontologies can be reduced to weakly linear programs and that query answering over such ontologies using a datalog engine is feasible in practice. We have evaluated our techniques on a large ontology repository. Our results show that many non-Horn ontologies can be rewritten into WL programs, and thus into datalog. We have tested the scalability of query answering using our approach, with promising results.
Researcher Affiliation Academia Mark Kaminski and Yavor Nenov and Bernardo Cuenca Grau Department of Computer Science, University of Oxford, UK
Pseudocode Yes Procedure 1 Rewrite Input: P: a disjunctive program Output: a datalog rewriting of P
Open Source Code No The paper does not provide a direct link or explicit statement about the release of source code for the methodology described. It mentions that modified ontologies can be found online, but not the code for their rewriting procedure.
Open Datasets Yes UOBM (Ma et al. 2006) is a standard benchmark for which synthetic data is available (Zhou et al. 2013). We denote the dataset for k universities by Uk. DBpedia is a realistic ontology with a large dataset from Wikipedia.
Dataset Splits No The paper does not specify explicit training/validation/test splits, percentages, or sample counts for their experiments. It describes using existing datasets (UOBM, DBpedia) for query answering without detailing how these were partitioned for different phases of model development.
Hardware Specification Yes We used a server with two Intel Xeon E5-2643 processors and 128GB RAM.
Software Dependencies No The paper mentions using RDFox, HermiT, Pellet, and KAON2 but does not provide specific version numbers for these software tools, which is necessary for reproducibility.
Experiment Setup Yes Timeouts were 10min for one query and 30min for all queries; a limit of 100GB was allocated to each task. We ran RDFox on 16 threads.