OWL

Web Ontology Language (OWL) extends RDFS syntax and formalises its semantics in terms of Description Logics. It makes richer models of knowledge about things possible, but at the cost of those models being more complex for a computer to process.

OWL is an ontology language which is logically sound and complete and at the same time implementable. It allows for more advanced inferences than over RDF with RDFS. OWL is a logical knowledge model which uses RDF/XML syntax for representaton.

There are thee flavours of OWL: OWL-Lite, OWL-DL, OWL-Full. OWL-Lite is a simplified subset of OWL-DL, and OWL-DL is a subset of OWL-Full. OWL-Lite and OWL-DL build on a restricted subset of RDFS terms. RDF-Full fully supports RDFS, but the ontologies produced in this dialect of OWL are computationally intractable.

OWL-Lite

OWL has the concepts of class, property, and instance.

OWL-Lite is its simplest dialect. It allows to denote equality and inequality between classes and properties with the following predicates:

equivalentClass
equivalentProperty
sameAs
differentFrom
AllDifferent
distinctMembers

These predicates give rise to particular inferences. For instance, the equivalentClass predicate asserts that two classes are equivalent. This means that it will be possible to infer that an instance of one class is also an instance of another class and vice versa. If Office and Bureau are equivalent classes, e.g.

Office equivalentClass Bureau

and CNN Office is declared an instance of the class Office, e.g.

CNN Office rdf:type Office

it will be possible to infer that it is also an instance of the class Bureau, e.g.

CNN Office rdf:type Bureau

OWL-Lite defines property characteristics, such as inverseOf, transitiveProperty, SymmetricProperty, FunctionalProperty, InverseFunctionalProperty. The property characteristics denote particular relationships between the elements of the subject predicate object triples. They provide mechanisms for infering new knowledge. For instance, the predicate ChildOf is defined as inverseOf ParentOf, e.g.

ChildOf inverseOf ParentOf

This enables the following inference: the subject of the first predicate is the object of the second predicate, and the object of the first predicate is the subject of the second predicate. Then it can be infered from the statement:

Maria isChildOf Peter

that

Peter isParentOf Maria

OWL-Lite defines also property restrictions, and restricted cardinality. The property restrictions allValuesFrom and someValuesFrom indicate how properties can be used by instances of a class. For instance, if the property subject Person of the property isMotherOf is restricted with allValuesFrom with the value Woman, it will be possible to infer from the statement:

Maria isMotherOf Harry

that Maria is a Woman , e.g.

Maria rdf:type Woman

The cardinality restrictions of OWL-Lite are minCardinality, maxCardinality and cardinality. They are stated on properties with respect to a particular class. The restrictions constrain the cardinality of that property on instances of that class. For instance, the class Child will have a cardinality restriction minCardinality of 1 on the property hasParent. If it is known that Georgia is of class Child, it will be possible to infer that Georgia has at least one parent.

OWL Lite cardinality restrictions allow only statements concerning cardinalities of value 0 or 1. OWL-DL and OWL Full allow arbitrary values for cardinality.

OWL-DL

OWL-DL is an extension of OWL-Lite. It provides with mechanisms to: Class construction enables the boolean combinations intersection, union, complement. OWL DL allows arbitrary combinations of classes with these restrictions. New classes are constructed based on the relationships intersectionOf, unionOf, complementOf between two classes. For instance, the class of University can be constructed from the intersection of the Class Academic Insitutions and the Class Educational Institutions, e.g.

intersectionOf(Academic Institutions, Educational Institutions)

The class of Bankers or Financial Advisers can be constructed from the union of the classes Bankers and Financial Advisers, e.g.

unionOf(Bankers, Financial Advisers)

And the class of Child can be constructed as a complement of the class Human Adults, e.g.

complementOf(Human Adults)

Class construction enables the building of a class out of the enumeration of named individuals which are complete extension of this class. This is done by using the predicate oneOf. For instance, the class of persons working for a company is described with this predicate, e.g.

oneOf(Bill, Allan, George, Venus, ...)

Class construction enables the building of a class from the restrictions of properties with respect to the class they apply to. The defined restrictions are someValuesFrom, allValuesFrom and hasValue. For instance, the class Manager can be restricted in a way that it includes only employees who are Engineers, e.g.

allValuesFrom(hasEmployee, Engineer)

Expressing the restrictions in OWL DL requires: (a) defining a class of type owl:Restriction, (b) refering to the property which will be restricted, and (c) defining the restriction itself.

OWL DL allows the definition of two disjoint classes. For instance, the class of Man and the class of Woman, e.g.

disjointWith(Man, Woman)

Property construction enables the definition of Object Properties and DataTypeProperties. They determine what kind of resources can be in the domain and in the range of the property - objects or datatypes. For instance, the property governs will have objectProperty with domain Government, and range Country, e.g.

govern rdf:type ObjectProperty;
rdfs:domain Government;
rdfs:range Country.

The property hasAge will have as range a DatatypeProperty - integer.

Class axioms define sub-class relationships, equivalent relationships and disjoint classes.

Property axioms define property equivalence and disjointness, and other property characteristics, which give more expressivity in the inference of characteristics pertaining to instances and their properties. Such characteristics are transitivity, symmetry, functional and inverse. They define particular relationships between the subjects and the objects of the properties, and enable inference of particular facts.

Individual axioms define relationships between individuals. It is possible to express that two individuals are the same or different with the properties sameAs, differentFrom and allDifferent. FOr instance, it is possible to express that Jacko and Michael Jackson are the same person using the predicate sameAs, e.g.

sameAs(Jacko, Michael Jackson)

Owl DL individuals are all implicitly members of the class owl:Thing. There is also an OWL DL class of which no individuals are members, e.g. owl:Nothing.

OWL DL has rich inference capabilities over ontologies. They rely on consistency and classification. For consistancy the correctness of an ontology is determined by calculating whether all statements can be satisfied. For classification the classes to which an individual belongs can be determined by showing that the individual satisfies the relevant conditions for class membership. Classes are defined in terms of necessary and sufficient conditions. Necessary conditions are conditions that must hold if an individual is a member of a class, e.g. property constraints. Sufficient conditions are conditions that mean an individual must be a member of a class, e.g. a superclass.

OWL-Full

Owl-Full is a superset of Owl DL. They both use the same vocabulary. Owl-Full includes all features of RDFS which can make ontologies computationally undecidable. In practice, OWL-Full is used where the ability to express knowledge is more important than the inferencing over it. Often OWL-Full is used to model full knowledge expression, and OWL-DL is used for partial knowledge inference.

OWL is a W3C recommendation.


Copyright © 2008-2009 Ontotext AD