This is the companion web page for our paper titled "R2O2: An Efficient Ranking-Based Reasoner for OWL Ontologies" accepted at ISWC 2015. If you have any questions regarding the research work or software, please feel free to contact me.
The ontologies we used in this paper are collected from the ORE 2014 reasoner competition. There are a total of 16,555 ontologies, and we selected 25% randomly. In total 4,138 ontologies were used in the paper.
The metric values of these ontologies can be found in subset1.csv.
Reasoner performance of the task of ontology classification was measured, and the wall time of each reasoner is recorded. The following csv files record the metric values and runtime for all the ontologies and for all the 6 reasoners.
- FaCT++: subset1_FACT.csv
- HermiT: subset1_HERMIT.csv
- HermiT: subset1_JFACT.csv
- Pellet: subset1_KONCLUDE.csv
- HermiT: subset1_MORE.csv
- TrOWL: subset1_TROWL.csv
Note that a 3,600-second (one hour) cut-off time was applied on all reasoners.
Meta reasoner R2O2
We employed 10-fold cross validation to assess the performance of R2O2, where two versions of R2O2 were evalated: with and without TrOWL.
The zip file contains two sets of the data we used in our evaluation. Once extracted, there will be two directories:
- 5_reasoners/: this directory contains the 10-fold datasets used to build R2O2 using 5 reasoners without TrOWL: FaCT++, Hermit, JFaCT, Konclude, and More.
- 6_reasoners/: this directory contains the 10-fold datasets used to build R2O2 using all of the 6 reasoners, including TrOWL.
In each of the 10 folds, the following data is provided:
- Actual reasoning performance in the training data combined with their ontology metrics used to build a ranking matrix: The name of each reasoner's reasoning performance data is seen as "r.reasonername_reg.csv". (e.g. FaCT++'s data: "r.fact_reg.csv")
- The ranking matrix built from the above reasoning performance data: "rank_train.csv"
- The random forest-based regression models developed for the reasoners developed in R: "reasoners_*.rdata"
- Testing data
- Actual reasoning performance of the testing data: "meta-test-time-actual.csv"
- Predicted reasoning performance of the testing data developed by the above the random-forest prediction models for the reasoners: "meta-test-time-predicted.csv"
- Actual ranking of reasoners according to their actual reasoning performance: "meta-test-actual.csv"
- Predicted ranking of reasoners according to their predicted reasoning performance: "meta-test-predicted.csv"