private static void print(Node<OWLClass> parent, OWLReasoner reasoner, int depth) { // We don't want to print out the bottom node (containing owl:Nothing and unsatisfiable classes) // because this would appear as a leaf node everywhere if (parent.isBottomNode()) { return; } // Print an indent to denote parent-child relationships printIndent(depth); // Now print the node (containing the child classes) printNode(parent); for (Node<OWLClass> child : reasoner.getSubClasses(parent.getRepresentativeElement(), true)) { // Recurse to do the children. Note that we don't have to worry about cycles as there // are non in the inferred class hierarchy graph - a cycle gets collapsed into a single // node since each class in the cycle is equivalent. print(child, reasoner, depth + 1); } }
public static void main(String[] args) { try { // Create our ontology manager in the usual way. OWLOntologyManager manager = OWLManager.createOWLOntologyManager(); // Load a copy of the people+pets ontology. We'll load the ontology from the web (it's // acutally located // in the TONES ontology repository). IRI docIRI = IRI.create(DOCUMENT_IRI); // We load the ontology from a document - our IRI points to it directly OWLOntology ont = manager.loadOntologyFromOntologyDocument(docIRI); System.out.println("Loaded " + ont.getOntologyID()); // We need to create an instance of OWLReasoner. An OWLReasoner provides the basic // query functionality that we need, for example the ability obtain the subclasses // of a class etc. To do this we use a reasoner factory. // Create a reasoner factory. In this case, we will use HermiT, but we could also // use FaCT++ (http://code.google.com/p/factplusplus/) or // Pellet(http://clarkparsia.com/pellet) // Note that (as of 03 Feb 2010) FaCT++ and Pellet OWL API 3.0.0 compatible libraries are // expected to be available in the near future). // For now, we'll use HermiT // HermiT can be downloaded from http://hermit-reasoner.com // Make sure you get the HermiT library and add it to your class path. You can then // instantiate the HermiT reasoner factory: // Comment out the first line below and uncomment the second line below to instantiate // the HermiT reasoner factory. You'll also need to import the // org.semanticweb.HermiT.Reasoner // package. OWLReasonerFactory reasonerFactory = null; // OWLReasonerFactory reasonerFactory = new Reasoner.ReasonerFactory(); // We'll now create an instance of an OWLReasoner (the implementation being provided by HermiT // as // we're using the HermiT reasoner factory). The are two categories of reasoner, Buffering // and // NonBuffering. In our case, we'll create the buffering reasoner, which is the default kind // of reasoner. // We'll also attach a progress monitor to the reasoner. To do this we set up a configuration // that // knows about a progress monitor. // Create a console progress monitor. This will print the reasoner progress out to the // console. ConsoleProgressMonitor progressMonitor = new ConsoleProgressMonitor(); // Specify the progress monitor via a configuration. We could also specify other setup // parameters in // the configuration, and different reasoners may accept their own defined parameters this // way. OWLReasonerConfiguration config = new SimpleConfiguration(progressMonitor); // Create a reasoner that will reason over our ontology and its imports closure. Pass in the // configuration. OWLReasoner reasoner = reasonerFactory.createReasoner(ont, config); // Ask the reasoner to do all the necessary work now reasoner.precomputeInferences(); // We can determine if the ontology is actually consistent (in this case, it should be). boolean consistent = reasoner.isConsistent(); System.out.println("Consistent: " + consistent); System.out.println("\n"); // We can easily get a list of unsatisfiable classes. (A class is unsatisfiable if it // can't possibly have any instances). Note that the getUnsatisfiableClasses method // is really just a convenience method for obtaining the classes that are equivalent // to owl:Nothing. In our case there should be just one unsatisfiable class - "mad_cow" // We ask the reasoner for the unsatisfiable classes, which returns the bottom node // in the class hierarchy (an unsatisfiable class is a subclass of every class). Node<OWLClass> bottomNode = reasoner.getUnsatisfiableClasses(); // This node contains owl:Nothing and all the classes that are equivalent to owl:Nothing - // i.e. the unsatisfiable classes. // We just want to print out the unsatisfiable classes excluding owl:Nothing, and we can // used a convenience method on the node to get these Set<OWLClass> unsatisfiable = bottomNode.getEntitiesMinusBottom(); if (!unsatisfiable.isEmpty()) { System.out.println("The following classes are unsatisfiable: "); for (OWLClass cls : unsatisfiable) { System.out.println(" " + cls); } } else { System.out.println("There are no unsatisfiable classes"); } System.out.println("\n"); // Now we want to query the reasoner for all descendants of vegetarian. Vegetarians are // defined in the // ontology to be animals that don't eat animals or parts of animals. OWLDataFactory fac = manager.getOWLDataFactory(); // Get a reference to the vegetarian class so that we can as the reasoner about it. // The full IRI of this class happens to be: // <http://owl.man.ac.uk/2005/07/sssw/people#vegetarian> OWLClass vegPizza = fac.getOWLClass(IRI.create("http://owl.man.ac.uk/2005/07/sssw/people#vegetarian")); // Now use the reasoner to obtain the subclasses of vegetarian. // We can ask for the direct subclasses of vegetarian or all of the (proper) subclasses of // vegetarian. // In this case we just want the direct ones (which we specify by the "true" flag). NodeSet<OWLClass> subClses = reasoner.getSubClasses(vegPizza, true); // The reasoner returns a NodeSet, which represents a set of Nodes. // Each node in the set represents a subclass of vegetarian pizza. A node of classes contains // classes, // where each class in the node is equivalent. For example, if we asked for the // subclasses of some class A and got back a NodeSet containing two nodes {B, C} and {D}, then // A would have // two proper subclasses. One of these subclasses would be equivalent to the class D, and the // other would // be the class that is equivalent to class B and class C. // In this case, we don't particularly care about the equivalences, so we will flatten this // set of sets and print the result Set<OWLClass> clses = subClses.getFlattened(); System.out.println("Subclasses of vegetarian: "); for (OWLClass cls : clses) { System.out.println(" " + cls); } System.out.println("\n"); // In this case, we should find that the classes, cow, sheep and giraffe are vegetarian. Note // that in this // ontology only the class cow had been stated to be a subclass of vegetarian. The fact that // sheep and // giraffe are subclasses of vegetarian was implicit in the ontology (through other things we // had said) // and this illustrates why it is important to use a reasoner for querying an ontology. // We can easily retrieve the instances of a class. In this example we'll obtain the // instances of // the class pet. This class has a full IRI of <http://owl.man.ac.uk/2005/07/sssw/people#pet> // We need to obtain a reference to this class so that we can ask the reasoner about it. OWLClass country = fac.getOWLClass(IRI.create("http://owl.man.ac.uk/2005/07/sssw/people#pet")); // Ask the reasoner for the instances of pet NodeSet<OWLNamedIndividual> individualsNodeSet = reasoner.getInstances(country, true); // The reasoner returns a NodeSet again. This time the NodeSet contains individuals. // Again, we just want the individuals, so get a flattened set. Set<OWLNamedIndividual> individuals = individualsNodeSet.getFlattened(); System.out.println("Instances of pet: "); for (OWLNamedIndividual ind : individuals) { System.out.println(" " + ind); } System.out.println("\n"); // Again, it's worth noting that not all of the individuals that are returned were explicitly // stated // to be pets. // Finally, we can ask for the property values (property assertions in OWL speak) for a given // individual // and property. // Let's get the property values for the individual Mick, the full IRI of which is // <http://owl.man.ac.uk/2005/07/sssw/people#Mick> // Get a reference to the individual Mick OWLNamedIndividual mick = fac.getOWLNamedIndividual(IRI.create("http://owl.man.ac.uk/2005/07/sssw/people#Mick")); // Let's get the pets of Mick // Get hold of the has_pet property which has a full IRI of // <http://owl.man.ac.uk/2005/07/sssw/people#has_pet> OWLObjectProperty hasPet = fac.getOWLObjectProperty(IRI.create("http://owl.man.ac.uk/2005/07/sssw/people#has_pet")); // Now ask the reasoner for the has_pet property values for Mick NodeSet<OWLNamedIndividual> petValuesNodeSet = reasoner.getObjectPropertyValues(mick, hasPet); Set<OWLNamedIndividual> values = petValuesNodeSet.getFlattened(); System.out.println("The has_pet property values for Mick are: "); for (OWLNamedIndividual ind : values) { System.out.println(" " + ind); } // Notice that Mick has a pet Rex, which wasn't asserted in the ontology. // Finally, let's print out the class hierarchy. // Get hold of the top node in the class hierarchy (containing owl:Thing) // Now print the hierarchy out Node<OWLClass> topNode = reasoner.getTopClassNode(); print(topNode, reasoner, 0); } catch (UnsupportedOperationException exception) { System.out.println("Unsupported reasoner operation."); } catch (OWLOntologyCreationException e) { System.out.println("Could not load the pizza ontology: " + e.getMessage()); } }