OWL API：如何确保域和范围限制？

Question

我尝试使用OWL API 来创建OWL Ontologies。我可以定义类、个体和它们之间的关系。

当我用域#A 和范围#B 定义对象属性#hasPart 时，我希望这个属性只能应用于这两个类的个体。但实际上 API 并不关心限制，因此也可以在 #C 类的两个成员之间分配 #hasPart，例如：

import org.semanticweb.owlapi.apibinding.OWLManager;
import org.semanticweb.owlapi.model.*;

public class OwlTest
{
    public static void main(String[] args)
    throws org.semanticweb.owlapi.model.OWLOntologyStorageException, org.semanticweb.owlapi.model.OWLOntologyCreationException, Exception
    {
        OWLOntologyManager manager = OWLManager.createOWLOntologyManager();
        OWLDataFactory df = manager.getOWLDataFactory();
        OWLOntology o = manager.createOntology();

       //------------------------------------------------------------------

        OWLClass clsA = df.getOWLClass( IRI.create("#A") );
        OWLClass clsB = df.getOWLClass( IRI.create("#B") );
        OWLClass clsC = df.getOWLClass( IRI.create("#C") );

        OWLObjectProperty hasPart = df.getOWLObjectProperty( IRI.create("#hasPart") );
        OWLObjectPropertyDomainAxiom domainAxiom = df.getOWLObjectPropertyDomainAxiom(hasPart, clsA);
        OWLObjectPropertyRangeAxiom   rangeAxiom = df.getOWLObjectPropertyRangeAxiom( hasPart, clsB);

        manager.addAxiom(o, domainAxiom);
        manager.addAxiom(o,  rangeAxiom);

       //------------------------------------------------------------------

        OWLNamedIndividual a1 = df.getOWLNamedIndividual( IRI.create("a1") );
        OWLNamedIndividual b1 = df.getOWLNamedIndividual( IRI.create("b1") );
        OWLNamedIndividual c1 = df.getOWLNamedIndividual( IRI.create("c1") );
        OWLNamedIndividual c2 = df.getOWLNamedIndividual( IRI.create("c2") );

        manager.addAxiom(o, df.getOWLClassAssertionAxiom(clsA, a1));
        manager.addAxiom(o, df.getOWLClassAssertionAxiom(clsB, b1));
        manager.addAxiom(o, df.getOWLClassAssertionAxiom(clsC, c1));
        manager.addAxiom(o, df.getOWLClassAssertionAxiom(clsC, c2));

        manager.addAxiom(o, df.getOWLObjectPropertyAssertionAxiom(hasPart, c1, c2));   // ObjectProperty '#hasPart' should only work for objects from Domain 'clsA' and Range 'clsB'

       //------------------------------------------------------------------

        manager.saveOntology(o, IRI.create("file:/tmp/data.owl"));
    }
}

输出/tmp/data.owl:

...
    <ObjectProperty rdf:about="#hasPart">
        <rdfs:domain rdf:resource="#A"/>
        <rdfs:range rdf:resource="#B"/>
    </ObjectProperty>

    <Class rdf:about="#A"/>
    <Class rdf:about="#B"/>
    <Class rdf:about="#C"/>

    <NamedIndividual rdf:about="a1">
        <rdf:type rdf:resource="#A"/>
    </NamedIndividual>

    <NamedIndividual rdf:about="b1">
        <rdf:type rdf:resource="#B"/>
    </NamedIndividual>

    <NamedIndividual rdf:about="c1">
        <rdf:type rdf:resource="#C"/>
        <p1:hasPart rdf:resource="c2"/>
    </NamedIndividual>
...

我现在正在寻找以编程方式处理此类限制的推荐方法..？非常感谢！

Answer 1

是的，使用 hasPart w/C 并没有错，推理者只会假设您最终会告诉它 c1 也是 A，或者 C 与 A 相同。

OWL-API 不会强制执行您想要的行为，听起来您正在寻找某种完整性约束，就像在正常的关系系统中一样。您要么必须将其烘焙到您的应用程序中，要么查看 Pellet 的 integrity constraints 之类的东西，它将在即将发布的 Pellet 3 版本中提供，目前在 Stardog 中提供。