OWL research at the University of Manchester

Joint research by members of the Information Management Group and the Bio-Health Informatics Group.

Supporting material: Subsumption Test Hardness in DL Classification

cade2014_subtestSupporting material for submission: Empirical Investigation of Subsumption Test Hardness in Description Logic Classification, as submitted to CADE-25, 2015.


The BioPortal Snapshot used in this research can be downloaded here:
10.5281/zenodo.15667.


The full set of modules used in the in-depth characterisation (experiment 2) can be downloaded here:
10.5281/zenodo.15682


The dataset is comprised of modules of the following ontologies:

MFOEM – Emotion Ontology
BT – Biotop Ontology
NEMO – Neural Electromagnetic Ontology
NTDO – Neglected Tropical Disease Ontology
OBI – Ontology for Biomedical Investigations
OGSF – Ontology for Genetic Susceptibility Factor
ONL-MSA – Mental State Assemssment
VSO – Vital Sign Ontology
CAO – Clusters of Orthologous Groups Cog Analysis Ontology
ICO – Informed Consent Ontology
OMRSE – Ontology of Medically Related Social Entities
STATO – Statistics Ontology
NPO – Nanoparticle Ontology
OBCS – Ontology of Biological and Clinical Statistics


The data driving the analysis along with the relevant R Scripts can be obtained from here:
10.5281/zenodo.15668


One interesting plot based on the subsumption test hardness survey showing how different reasoners compare by ontology can be downloaded here.


One large grid plot comparing the categories subsumption tests fall into in the inter-module analysis, arranged in a grid by ontologies and reasoners, can be downloaded here.


The following table shows where the different measures in the OWL Reasoner Stage Benchmark are taken. PP (Pre-processing) indicates the start of the process and is taken before OWLReasonerFactory.createReasoner() is called. POP_FIN (Postprocessing finish) marks the end of the classification process and is taken right after OWLReasoner.precomputeInferences(CLASS_HIERARCHY) (or the respective equivalent).

JFact
PP OUTSIDE
CC TBox.isConsistent();
PRP TBox.isConsistent()
ST TBox.createTaxonomy()
POP TBox.createTaxonomy()
POPFIN OUTSIDE
SAT TBox.isSubsHold(Concept,Concept)
   
  HermiT
PP OUTSIDE
CC Reasoner.isConsistent()
PRP Reasoner.isConsistent()
ST DeterministicClassification.classify(); QuasiOrderClassification.buildHierarchy();
POP DeterministicClassification.classify(); QuasiOrderClassification.buildHierarchy();
POPFIN OUTSIDE
SAT Tableau.isSatisfiable(boolean loadPermanentABox,boolean loadAdditionalABox,Set perTestPositiveFactsNoDependency,Set perTestNegativeFactsNoDependency,Set perTestPositiveFactsDummyDependency,Set perTestNegativeFactsDummyDependency,Map termsToNodes,Map nodesForIndividuals,ReasoningTaskDescription reasoningTaskDescription)
   
  Pellet
PP OUTSIDE
CC KnowledgeBase.consistency()
PRP KnowledgeBase.consistency()
ST KnowledgeBase.classify()
POP KnowlegeBase.classify()
POPFIN OUTSIDE
SAT ABox.isSatisfiable(ATermAppl c1, ATermAppl c2);