Supplementary material of the experiments with rough ontologies


This page contains additional files for experiments 1 & 2 and 3, and the test files for the subsumption reasoning searvices.

Experiment 1 & 2: promiscuous bacteriain OWL 2 DL and OWL 2 QL

  • A graphical depiction of a section of the HGT ontology, with the rough concept "PromiscuousBacterium", and as owl file for the OBDA, which is set up as described earlier on this page.
  • The updated mapping file that contains the SQL queries to retrieve the promiscuous bacteria from the HGT-DB database. A nicer GUI rendering of the mapping in the OBDA plugin for Protégé is shown in the figure below.
  • The exported query answer to retrieve the instances of PromiscuousBacterium where the indistinguishable bacteria are denoted with their abbreviation dehaloc (Dehalococcoides CBDB1) and tmar (Thermotoga maritima),
  • A refinement of the rough concept, PromBactPrime, that turns out to have no indistinguishable bacteria.
  • The more comprehensive semantics (w.r.t. rough sets) of the promiscuous bacterium, its lower and upper approximation, and the equivalence relation (ind) in the OWL 2 DL version of the ontology (more precisely, in SRIQ(D)). The first figure shows part of the decsription of Promiscuousbacterium, and the second one of its uppoer approximation.
There is also more information about the HGT-DB and its use with the Ontology-Based Data Access framework, including the SAC'10 paper about the WONDER system.

Experiment 3: septic patients

  • Informal definitions from Schlobach et al's paper for the Bone criteria: and the corresponding Bone Criteria and Strict Septic definitions how it should be in Protege.
  • OWL files: roughtestIND.owl with TBox and ABox that lets Protege quit 'unexpectedly', and two other versions, one without RoughC subsumptions, one with. Observe that Arturo should be classified as an instance of StrictSeptic, based on the values and the definitions.
  • Given rough sets' semantics, both BoneSeptic and StrictSeptic each can be a rough concept (instead of being upper and lower approximation, respectively). This is modelled as such in the rough sepsis owl file. Note that also here, upon copying the right definition (see bullet point 1) in the owl file, Protege quits unexpectedly upon classifying. One can copy it in, save, let it quit unexpectedly, re-open so that the definition has been modified, and then run classify with FaCT++; this works, but will still misclassify Arturo.
  • RacerPro with two versions of the definitions in the t-box, being one variation with "urine OR dialysis" and one with "urine AND dialysis", and the a-box assertions.

(Rough) Subsumption reasoning