Web Ontology Language (OWL):
Guide Version 1.0

This version:
29 October 2002
Previous version:
24 Oct 2002
Michael K. Smith, Electronic Data Systems, michael.smith@eds.com
Raphael Volz, Forschungszentrum Informatik (FZI), volz@fzi.de
Deborah McGuinness, Stanford University, dlm@ksl.stanford.edu
This document represents the work of many people, in particular the members of the W3C Web Ontology Working Group. Appendix B was contributed by Guus Schreiber, University of Amsterdam, schreiber@swi.psy.uva.nl. Substantial insight was provided by the DAML+OIL Walkthru. Jeremy Carroll, Jeff Heflin and Leo Obrst provided extensive reviews. At the WG Face to Face, October 8, 2002, Stephen Buswell, Ruediger Klein, Enrico Motta, Evan Wallace, and Chris Welty provided a detailed review of the ontology resulting in substantial changes.


The World Wide Web as it is currently constituted resembles a poorly mapped geography. Our insight into the documents and capabilities available are based on keyword searches, abetted by clever use of document connectivity and usage patterns. The sheer mass of this data is unmanageable without powerful tool support. In order to map this terrain more precisely, computational agents require machine-readable descriptions of the content and capabilities of web accessible resources. These descriptions must be in addition to the human-readable versions of that information.

The Web Ontology Language (OWL) is intended to provide a language that can be used to describe the classes and relations between them that are inherent in Web documents and applications.

This document demonstrates the use of the OWL language to

  1. formalize a domain by defining classes and properties of those classes,
  2. define individuals and assert properties about them, and
  3. reason about these concepts to the degree permitted by the formal semantics of the OWL language.
The sections are organized to present an incremental definition of a set of classes beginning with the fundamentals and proceeding to more complex language components.



"Tell me what wines I should buy to serve with the following menu. And, by the way, I don't like Sauterne."

It would be difficult today to construct a Web agent that would be capable of performing a search for wines on the web that satisfied this query. Similarly, consider actually assigning a software agent the task of making a coherent set of travel arrangements. (For more use cases see the OWL requirements document.)

To support this sort of computation, it is necessary to go beyond keywords and specify the meaning of the resources described on the web. This additional layer of interpretation captures the semantics of the data.

OWL provides a language to define Web ontologies. An ontology describes the concepts in a specific domain of discourse. In the case of OWL, an ontology may include

Datatype properties and object properties are collectively the properties of a class.

Defined classes and properties can then be used to describe specific Web resources. A great deal of information may be available about an object based on a simple categorization. Due to the properties of the class and its various parent classes, many additional derived characteristics of the instance may be entailed (logically implied) by the semantics of OWL.

One question that comes up when describing yet another XML/Web standard is "What does this buy me that XML and XML Schema don't?" An operational consensus can always be developed over the meaning of a set of XML tags and their contents. There is furious ongoing standards activity doing exactly this.

There are two answers to this question.


There are several ways to think of OWL, depending on your background.

  1. Object oriented programming: Consider classes to be object definitions, with functional properties as fields and relational properties as boolean methods.
  2. Set theory: This maps very naturally onto the semantics. Classes define sets of individuals and a property can be taken to be the set of pairs for which it is true.
  3. Logical predicates: Think of classes as one-place predicates defining types, where C(x) means that x is an element of class C. And let properties be two-place predicates.

These are approximations. To understand OWL, read on.

Structure of the Document

In order to provide a consistent set of examples throughout the guide, we have created a wine ontology based on an earlier DAML version that was developed as a description logic example and tutorial.

In this document we present examples using the OWL XML syntax, assuming XML will be familiar to the largest audience. The standard for interchange of OWL assertions between tools depends on RDF triples. These XML and RDF formats are part of the standard. Other notations have been formulated, in particular a UML version. Appendix A provides links to various primers on related standards.

Notes indicating expected changes pending clarification are marked with '@@', e.g. [@@ Note]. All of the examples presented in this document are taken from the ontologies contained in wine.owl and food.owl, except those marked with  ¬  in the bottom right corner.

The Structure of Ontologies

We consider an ontology to be a specification or model for a particular domain or set of domains. Given a domain, an ontology defines its basic and derived concepts and their interrelationships. As we use the term, an ontology will often include assertions about individuals in addition to classes. That is, the ontology might include a class, Dog, as well as an instance, Dog45, identifying the particular dog M. Smith owned from 1986 through 2000. The property Name may be defined as a relation between Dogs and strings, while an instance of the Name property would exist that connects Dog45 with the specific value, "Tuxedo".

In order to write down an ontology that can be interpreted unambiguously and used by software agents we require a formal syntax and semantics. OWL provides these necessary underpinnings for ontology construction. See the formal semantics and XML syntax.

A set of OWL assertions loaded into a reasoning system is called a knowledge base (KB). These assertions may be based on a single ontology or multiple distributed ontologies that have been combined using defined OWL mechanisms. The assertions may include various derived facts, facts not literally present in the original textual representation of the ontology, but entailed by the semantics of OWL.

The Species of OWL

OWL is in fact a set of three, increasingly complex languages.

The Distribution of Ontologies

OWL is a component of the Semantic Web activity. This effort aims to make Web resources more readily accessible to automated processes by adding information about the resources that describe or provide Web content. Because the Semantic Web deals with web content, it is inherently distributed. We expect OWL ontologies to also be distributed. One consequence of this is that OWL generally assumes an open world assumption. That is, descriptions of resources are not confined to a single file or scope. While concept C1 may be defined originally in ontology O1, it can be extended in other ontologies. The consequences of these additional propositions about C1 are monotonic. New information cannot negate previous information. Facts and entailments can only be added, never deleted.

OWL does allow negative information to be explicitly stated. As a result, it is fairly simple to explicitly state contradictions. In addition, contradictions may be implicit. These properties of a knowledge base are something the designer of an ontology needs to be careful about. It is expected that tool support will help detect such cases.


Before we can use a set of terms, we need a precise indication of what specific vocabularies are being used. Where do the terms come from? A standard initial component of an ontology includes a set of namespace declarations enclosed in an opening rdf:RDF tag. These provide a means to unambiguously interpret identifiers and make the rest of the ontology presentation much more readable. A typical OWL ontology begins with a namespace declaration similar to the following.

    xmlns     ="http://www.example.org/wine.owl#" 
    xmlns:vin ="http://www.example.org/wine.owl#"       
    xmlns:owl ="http://www.w3.org/??/owl#"
    xmlns:rdf ="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
    xmlns:xsd ="http://www.w3.org/2000/10/XMLSchema#"
    xmlns:dte ="http://www.w3.org/2003/??/owl-guide-dt#" > 

The first two declarations identify the namespace associated with this ontology. The first makes it the default namespace, stating that unprefixed elements and empty URI references refer to the current ontology. The second identifies the namespace of the current ontology with the prefix vin:. The third identifies the namespace of the supporting food ontology with the prefix food:.

The fourth namespace declaration says that in this document, elements prefixed with owl: should be understood as referring to things drawn from the namespace called http://www.w3.org/2003/??/owl#. This is a conventional OWL declaration, required to introduce the OWL vocabulary.

OWL depends on constructs defined by RDF and RDFS. In this document, the rdf: prefix is understood to refer to things drawn from the namespace called http://www.w3.org/1999/02/22-rdf-syntax-ns#. The next two namespace declarations make similar statements about the RDF Schema (rdfs:) and XML Schema datatype (xsd:) namespaces.

This OWL document is related to a separate document containing XML Schema datatype definitions. The final declaration says that elements prefixed with dte: should be understood as referring to things drawn from the namespace called http://www.w3.org/2003/??/owl-guide-dt#, which contains the datatype definitions for this Guide document.

As an aid to writing down references to lengthy URLs it can often be useful to provide a set of entity definitions in a document type declaration (DOCTYPE) that precedes the ontology definitions. The names defined by the namespace declarations only have significance as parts of XML tags. Attribute values are not namespace sensitive. But in OWL we frequently reference ontology identifiers using attribute values. They can be written down in their fully expanded form, for example "http://www.example.org/owl/wine.owl#merlot". Alternatively, abbreviations can be defined using an ENTITY definition, for example:

<!DOCTYPE owl [
    <!ENTITY vin  "http://www.example.org/wine.owl#" >
    <!ENTITY food "http://www.example.org/food.owl#" > ]>
After this pair of ENTITY declarations, we could write the value "&vin;merlot" and it would expand to "http://www.example.org/owl/wine.owl#merlot".

Ontology Headers

Once namespaces are established we begin with an assertion that what follows is an OWL ontology.

<owl:Ontology rdf:about="http://www.example.org/wine.owl"> 

This assertion is formulaic. The about attribute will normally be the URL of the current file, indicating that the subject of this assertion is this document. If desired, the ontology may be given a name that is a URN and independent of a particular physical location.

[@@ Check URN statement. XMLBase?]

Having established the fact that we are defining an ontology and provided the necessary namespace mapping, a set of additional tags are provided to handle such critical housekeeping tasks as comments, version control and inclusion of other ontologies.

<owl:Ontology rdf:about="http://www.example.org/wine.owl"> 
  <rdfs:comment>An example OWL ontology</rdfs:comment>
      $Id: Guide.html,v 1.0 2002/10/28 16:12:00 mksmith $
  <owl:imports rdf:resource="http://www.w3.org/2003/??/food.owl"/> 

<rdfs:comment> provides the obvious needed capability to annotate an ontology.

<owl:versionInfo> is a standard tag intended to provide a consistent hook for version control systems working with ontologies. OWL does not specify the structure of the content.

<owl:imports> provides an include-style mechanism. Importing another ontology brings the entire set of definitions provided by that ontology into the knowledge base. In order to make best use of this imported ontology it would normally be coordinated with a namespace declaration. Notice the distinction between these two mechanisms. The OWL namespace declaration provides a convenient shorthand to help reference names defined in other OWL ontologies. Conceptually, owl:imports is provided to indicate your intention to include the assertions of the target ontology. These assertions define the meaning of the terms defined in that ontology, meanings that support reasoning about the terms.

Note that owl:imports may not always succeed. As you would expect when dealing with the Semantic Web, access to resources distributed across the Web may not always be possible. Tools will respond to this situation in an implementation defined manner.

OWL provides several other mechanisms to tie the current ontology and imported ontologies together (see ontology mapping).

One common set of additional tags that could reasonably be included here are the standard Dublin Core metadata tags. Examples include Title, Creator, Description, Publisher, and Date. The URI defining the Dublin Core namespace is 'http://purl.org/dc/elements/1.1/'.)

The ontology header definition is closed with the following tag.

This prelude is followed by the actual conceptual definitions that make up the ontology and is ultimately closed by

Basic Definitions

Simple Classes and Individuals

Most uses of an ontology depend ultimately upon the ability to reason about individuals. In order to do this in a useful fashion we need to have a mechanism to describe the classes that individuals belong to and the properties that they inherit by virtue of class membership. We can always assert specific properties about individuals, but much of the power of ontologies comes from class-based reasoning.

Sometimes we want to emphasize the distinction between a class as an object and a class as a set containing elements. We call the set of individuals that are members of a class the extension of the class.

Defining Simple Hierarchical Named Classes
class, subClassOf

The most basic concepts in an ontology are the roots of various taxonomic trees. Every instance in the OWL world is a member of owl:Thing. Thus each of these user-defined primitive classes is implicitly a subclass of owl:Thing. Root classes are defined by simply declaring a named class.

We first create the classes of Winery, Region, and ConsumableThing.

<owl:Class rdf:ID="Winery"/> 
<owl:Class rdf:ID="Region"/> 
<owl:Class rdf:ID="ConsumableThing"/> 

Note that we have only said that there exist classes that have been given these names, indicated by the 'ID=' syntax. Formally, we know almost nothing about these classes other than their existence, despite the use of familiar English terms as labels. And while the classes exist, they may have no members. For all we know at the moment, these classes might as well have been called Thing1, Thing2, and Thing3.

It is important to remember that definitions may be incremental and distributed. We will have more to say about Winery later.

The attribute/value syntax rdf:ID="Region" is used to first introduce a name, as part of its definition. This is the rdf:ID attribute that is like the familiar ID attribute defined by XML. Within this document, the concept of a wine region can now be referred to using rdf:resource="#Region". Other ontologies may reference this name using its complete form, "documentURI#Region" (see below).

Another form of reference uses the syntax rdf:about="#Region" to extend the definition of a resource. This use of the rdf:about="#x" syntax is a critical element in the creation of a distributed ontology. It permits the extension of the imported definition of x without modifying the original document and supports the incremental construction of a larger ontology.

It is now possible to refer to these defined concepts in other OWL constructs by their given identifier. For the first class, within this document, we can use the relative identifier, #Winery. Other documents may need to reference this class as well. The most reasonable way to do so is to provide a namespace definition that includes the defining document as a source:

<rdf:RDF xmlns:vin="http://www.w3.org/2003/??/owl-guide#"  ... / > 
Given this namespace definition, vin:Winery can be used to refer to the winery concept that we defined above. (Remember that namespaces are not really intended to identify actual documents. In the case of an imported ontology, they just need to line up with the namespace definition in that document. We often succumb to the natural inclination to make the namespace and document identical.) More literally, it is always possible to reference a resource using its full URI, http://www.w3.org/2003/??/owl-guide#winery.

The fundamental taxonomic constructor for classes is subclassOf. It is transitive. If X is a subclass of Y and Y a subclassOf Z then X is a subclassOf Z.

<owl:Class rdf:ID="PotableLiquid"> 
  <rdfs:subClassOf rdf:resource="#ConsumableThing" />

We define PotableLiquid (liquids suitable for drinking) to be a subclass of ConsumableThing.

In the world of web-based ontologies, both of these concepts should be defined in a separate ontology that would provide the basic building blocks for a wide variety of food and drink ontologies. And in fact this is what we have done. These elements are defined in the food ontology which is imported into the wine ontology. (See owl:import for details.) The food ontology includes a number of concepts, for example Food, EdibleThing, MealCourse, and Shellfish, that do not belong in a collection of wine facts, but must be connected to the wine vocabulary if we are going to perform useful reasoning. Food and wine are mutually dependent, in order to satisfy our need to identify wine/food matches.

A class definition has two parts: a name introduction or reference and a list of restrictions. Each of the immediate contained expressions in the class definition further restricts the members of the defined class. Class elements are members of the intersection of the restrictions. So far we have only seen examples that include a single restriction, forcing the new class to be a subclass of some other named class.

At this point it is possible to create a simple (and incomplete) definition for the class Wine. Wine is a PotableLiquid.

<owl:Class rdf:ID="Wine"> 
  <rdfs:subClassOf rdf:resource="#PotableLiquid"/> 
  <rdfs:label xml:lang="en">wine</rdfs:label> 
  <rdfs:label xml:lang="fr">vin</rdfs:label> 

The rdfs:label entry provides an optional human readable name for this concept. Presentation tools can make use of it. The "lang" attribute provides support for multiple languages. A label is like a comment and contributes nothing to the logical interpretation of an ontology.

Our wine definition is still very incomplete. We know nothing about wines except that they are things and potable liquids, but we have sufficient information to create and reason about instances.

Design for Use

There are important issues regarding when a concept is an instance and when it is a class. The fundamental driver is how the ontology is going to be used.

The ontology that we have adapted for use in this document had elected to make instances be particular types of wine produced by a particular winery. This does not capture a number of distinctions, for example that between wines made in different years.

We added a subclass of Wine, Vintage, which has a VintageYear property. But there are always going to be distinctions that could be made that a particular ontology has not.

Wineries produce different versions of a single variety of wine. For example, Page Mill Vineyard, in 2000, sold two chardonnay's, Chardonnay Bien Nacido Vineyard and Chardonnay Arey Listwin Vineyard. Of course we can always create PageMillWineryChardonnay-1 and PageMillWineryChardonnay-2. But we have no Name, Price, Prizes, or Description slots for wines. These are the sort of properties that might distinguish these two concepts. Under our current ontology they would appear identical.

The point of this digression is to note that the development of an ontology should be firmly driven by use cases.

Defining Individuals

In addition to classes, we want to be able to describe their members. We normally think of these as individuals in our universe of things. An individual is minimally introduced by declaring it to be a member of a class.

<Region rdf:ID="CentralCoastRegion" /> 

Note that the following is identical in meaning to the definition above.

<owl:Thing rdf:ID="CentralCoastRegion" /> 

<owl:Thing rdf:about="#CentralCoastRegion"> 
   <rdf:type rdf:resource="#Region"/> 
rdf:type is an RDF property that ties an individual to a class of which it is a member.

There are a couple of points to be made here. First, we have decided that CentralCoastRegion (a specific area) is an instance of a Region, the class containing all geographical regions.

Second, there is no requirement in the two-part example that the two statements need to be adjacent to one another, or even in the same file (though the names would need to be extended with a URI in such a case). We design Web ontologies to be distributed. They can be imported and augmented, creating derived ontologies.

In order to have available a few more basic definitions for the properties introduced in the next sections, we define a branch of the Grape taxonomy, with the lone leaf being the Cabernet Sauvignon varietal.

<owl:Class rdf:ID="Grape">

<owl:Class rdf:ID="Varietal">
  <rdfs:subClassOf rdf:resource="#Grape"/>

<Varietal rdf:ID="CabernetSauvignonGrape" />
This means that we have in some sense limited our ontology for specific purposes. Because it is itself an instance, we cannot create an instance of CabernetSauvignonGrape. If we needed to talk about individual grapes or specific bunches of grapes, we would have designed the ontology differently.

Simple Properties

Defining properties
ObjectProperty, DatatypeProperty, subPropertyOf,
domain, range

This world of classes and individuals would be pretty uninteresting if we could only define taxonomies. Properties let us assert general facts about the members of classes and specific facts about individuals.

A property is a binary relation. Two types of properties are distinguished:

When we define a property there are a number of ways to restrict the elements of the relation. The domain and range can be specified. The property can be defined to be a specialization (subproperty) of an existing property. More elaborate restrictions are possible and are described later.
<owl:ObjectProperty rdf:ID="MadeFromVarietal"> 
  <rdfs:domain rdf:resource="#Wine"/>
  <rdfs:range rdf:resource="#Varietal"/> 
In OWL, a sequence of statements without an explicit operator represents an implicit conjunction. They apply to their containing element. The property MadeFromVarietal has a domain of Wine and a range of Varietal. That is, it relates elements of the class Wine to elements of the class Varietal.

Properties, like classes, can be arranged in a hierarchy.

<owl:ObjectProperty rdf:ID="WineDescriptor" />

<owl:Class rdf:ID="WineColor">
  <rdfs:subClassOf rdf:resource="#WineDescriptor" />

<owl:ObjectProperty rdf:ID="hasWineDescriptor">
  <rdfs:domain rdf:resource="#Wine" />
  <rdfs:range  rdf:resource="#WineDescriptor" />

<owl:ObjectProperty rdf:ID="hasColor">
  <rdfs:subPropertyOf rdf:resource="#hasWineDescriptor" />
  <rdfs:range rdf:resource="#WineColor" />
WineDescriptor properties relate wines to their color and components of their taste, including sweetness, body, and flavor. hasColor is a subproperty of the hasWineDescriptor property, with its range further restricted to WineColor.

It is now possible to expand the definition of Wine to include the notion that a wine is an element of things that are made from at least one Varietal. As with property definitions, class definitions have multiple subparts that are implicitly conjoined.

First we introduce the locatedIn property, which relates things to the regions they are located in.

<owl:ObjectProperty rdf:ID="locatedIn">
  <owl:TransitiveProperty />
  <rdfs:domain rdf:resource="http://www.w3.org/@@/owl#Thing" />
  <rdfs:range rdf:resource="#Region" />
Notice how the domain and range of locatedIn are defined. The domain permits anything to be located in a region. And the transitive compostion of this relation essentially creates a tree of geographic subregions whose leaves can be anything.

Now we extend the definition of Wine.

<owl:Class rdf:ID="Wine"> 
  <rdfs:subClassOf rdf:resource="#PotableLiquid"/> 
      <owl:onProperty rdf:resource="#madeFromVarietal"/>
      <owl:onProperty rdf:resource="#locatedIn"/> 
The highlighted subclass restriction defines an un-named class that represents the set of things with at least one madeFromVarietal property. We call these anonymous classes [OWL/DL]. Inserting this subclass restriction in the Wine class definition body states that things that are wines are also members of this anonymous class. That is, every wine instance must participate in at least one madeFromVarietal relation. Additionally, every wine must come from at least one region. This cliché is presented in more detail in the section on restrictions.

Properties and Datatypes

We distinguish properties according to whether they relate resources to resources (object properties) or resources to datatypes (datatype properties). Datatype properties may range over strings or they may make use of simpleTypes defined in accordance with XML Schema datatypes.

Suppose we wanted to restrict vintage years to years after 1700. We would create several XML Schema datatype definitions in a separate file, such as http://www.w3.org/@@/owl-ex-dt.xsd , which contains:

<xsd:schema xmlns:xsd="http://www.w3.org/2001/XMLSchema#" 

  <xsd:simpleType name="year"> 
    <!-- year is an XMLS datatype based on integer --> 
    <xsd:restriction base="xsd:decimal"/> 
  <xsd:simpleType name="wineYear"> 
    <!-- wineYear is an XMLS datatype based on year --> 
    <!-- with the added restriction that values must be GEQ 1700 --> 
    <xsd:restriction base="year"> 
    <xsd:minInclusive value="1700"/>

'Year' is defined to be a subtype of the XML Schema simple type 'decimal'. 'WineYear' is a subtype of 'Year' that is restricted to decimal numbers greater than or equal to 1700. We can reference elements of this file in OWL property definitions and class restrictions.

[@@ RE syntax of datatype reference. Awaiting input from RDF Core.]

<owl:Class rdf:ID="WineYear" />

<owl:DataTypeProperty rdf:ID="yearValue">
  <rdfs:domain rdf:resource="#WineYear" />    
  <rdfs:range  xsi:type="dte:wineYear" /> 
The yearValue property relates WineYears to decimal numbers greater than or equal to 1700.

Properties of Individuals

First we create Region and Winery instances, and then we define our first wine instance, a Cabernet Sauvignon.

<CaliforniaRegion rdf:ID="SantaCruzMountainsRegion" /> 

<Winery rdf:ID="SantaCruzMountainVineyard" />

<CabernetSauvignon rdf:ID="SantaCruzMountainVineyardCabernetSauvignon" >
  <locatedIn   rdf:resource="#SantaCruzMountainsRegion"/>  
  <hasMaker    rdf:resource="#SantaCruzMountainVineyard" />   
This is still incomplete. There are other aspects of the wine flavor that are defined in the full ontology. But the pieces are falling together. We could begin reasoning about what menu items this wine might accompany. We know from the definition above that the Santa Cruz Mountain Vineyard makes it. Because it is a Cabernet Sauvignon (see wine.owl), we know it is a dry, red wine.

Property Characteristics

The next few sections describe the mechanisms used to further define properties. It is possible to specify property characteristics, which provides a powerful mechanism for enhanced reasoning about a property.


A property, P, tagged as transitive satisfies the following axiom:

P(x,y) and P(y,z) -> P(x,z)
The concept locatedIn is transitive.
<owl:ObjectProperty rdf:ID="locatedIn">
  <owl:TransitiveProperty />
  <rdfs:domain rdf:resource="&owl;Thing" />
  <rdfs:range rdf:resource="#Region" />

<Region rdf:ID="SantaCruzMountainsRegion">
  <locatedIn rdf:resource="#CaliforniaRegion" />

<Region rdf:ID="CaliforniaRegion">
  <locatedIn rdf:resource="#UsRegion" />
Because the SantaCruzMountainsRegion is locatedIn the CaliforniaRegion, then it must also be locatedIn the USRegion, since locatedIn is transitive.


A property, P, tagged as symmetric satisfies the following axiom:

P(x,y) iff P(y,x)
The property adjacentRegion is symmetric, while locatedIn is not.
<owl:ObjectProperty rdf:ID="adjacentRegion">
  <owl:SymmetricProperty />
  <rdfs:domain rdf:resource="#Region" />
  <rdfs:range rdf:resource="#Region" />

<Region rdf:ID="MendocinoRegion">
  <locatedIn rdf:resource="#CaliforniaRegion" />
  <adjacentRegion rdf:resource="#SonomaRegion" />
The MendocinoRegion is adjacent to the SonomaRegion and vice-versa. The MendocinoRegion is located in the CaliforniaRegion but not vice versa.


A property, P, tagged as functional satisfies the following axiom:

P(x,y) and P(x,z) -> y = z 
In our wine ontology, hasVintageYear is functional. A wine has a unique vintage year. That is, a given instance of Vintage can only be associated with a single year.
<owl:Class rdf:ID="VintageYear" />

<owl:ObjectProperty rdf:ID="hasVintageYear">
  <owl:FunctionalProperty />
  <rdfs:domain rdf:resource="#Vintage" />
  <rdfs:range  rdf:resource="#VintageYear" />


A property, P1, tagged as the InverseOf P2, satisfies the following axiom:

P1(x,y) iff P2(y,x)
Note that the syntax for InverseOf takes a property name as an argument.
<owl:ObjectProperty rdf:ID="hasMaker">
  <owl:FunctionalProperty />
<owl:ObjectProperty rdf:ID="producesWine">
  <owl:InverseOf rdf:resource="#hasMaker" />
Wines have makers, which in the definition of Wine are restricted to Winerys. Then each Winery produces the set of wines that identify it as maker.


A property, P, tagged as InverseFunctional satisfies the following axiom:

P(y,x) and P(z,x) -> y = z 
Notice that producesWine in the preceding section is inverse functional. The reason is that the inverse of a functional property must be inverse functional. We could have defined hasMaker and producesWine as follows and achieved the identical effect as the preceding example.
<owl:ObjectProperty rdf:ID="hasMaker" />
<owl:ObjectProperty rdf:ID="producesWine">
  <owl:InverseFunctional />
  <owl:InverseOf rdf:resource="#hasMaker" />
</owl:ObjectProperty>                                     ¬ 
Think of the elements of the range in an inverseFunctional property as defining a unique key in the database sense. Inversefunctional implies that the elements of the range provide a unique identifier for each element of the domain.

Property Restrictions

In addition to designating property characteristics, it is possible to further constrain the range of a property in specific contexts in a variety of ways.

allValuesFrom, someValuesFrom

We have already seen one way to restrict the types of the elements that make up a property. The mechanisms to date have been global in that they apply to all instances of the property. These next two, allValuesFrom and someValuesFrom, are local to their containing class definition.

<owl:Class rdf:ID="Wine">
  <rdfs:subClassOf rdf:resource="&food;PotableLiquid" />
      <owl:onProperty rdf:resource="#hasMaker" />
      <owl:allValuesFrom rdf:resource="#Winery" />

The maker of a Wine must be a Winery. The allValuesFrom restriction is on the hasMaker property of this Wine class only. Makers of Cheese are not constrained by this local restriction.

owl:someValuesFrom is similar, but less restrictive. If we replaced owl:allValuesFrom with owl:someValuesFrom in the example above, it would mean that at least one of the hasMaker instances of Wine must point to an instance of a Winery.

<owl:Class rdf:ID="Wine">
  <rdfs:subClassOf rdf:resource="&food;PotableLiquid" />
      <owl:onProperty rdf:resource="#hasMaker" />
      <owl:someValuesFrom rdf:resource="#Winery" />
</owl:Class>                                             ¬ 
The difference between the two formulations is the difference between a universal and existential quantification.

  allValuesFromFor all wines, if they have makers, all the makers are wineries.
  someValuesFromFor all wines, they have at least one maker that is a winery.

The first does not require a wine to have a maker. If it does have one or more, they must all be wineries. The second requires that there be at least one maker that is a winery.


We have already seen examples of cardinality constraints. To date, they have been assertions about minimum cardinality. Even more straight-forward is owl:cardinality, which permits the specification of exactly the number of elements in a relation. For example, we define Vintage to be a subclass of Wine with exactly one VintageYear.

<owl:Class rdf:ID="Vintage"> 
  <rdfs:subClassOf rdf:resource="#Wine"/> 
      <owl:onProperty rdf:resource="#VintageYear"/>  

Cardinality expressions with values limited to '0' or '1' are part of OWL/Lite. This permits the user to indicate 'at least one', 'no more than one', and 'exactly one'. Other integers are only permitted in OWL/DL. owl:maxCardinality can be used to specify an upper bound. In combination with owl:minCardinality it is possible to specify a range.

hasValue [OWL/DL]

hasValue allows us to define classes based on the existence of particular property values. Hence, an individual will be a member of such a class whenever at least one of its property values is equal to the hasValue resource.

<owl:Class rdf:ID="Burgundy">
      <owl:onProperty rdf:resource="#hasSugar" />
      <owl:hasValue rdf:resource="#Dry" />
Here we define Burgundy wine to be dry wines. That is, their hasSugar property must have at least one value that is equal to Dry.

hasValue can only be used within a property restriction. The owl:onProperty statement indicates the restricted property. As for allValuesFrom and someValuesFrom, this is a local restriction. It holds for hasSugar as applied to Burgundy.

Ontology Mapping

In order for ontologies to have the maximum impact, they need to be widely shared. In order to minimize the intellectual effort involved in developing an ontology they need to be re-used. In the best of all possible worlds they need to be composed. For example you might adopt a date ontology from one source and a physical location ontology from another.

It is important to realize that much of the effort of developing an ontology is devoted to hooking together classes and properties in ways that maximize implications. We want simple assertions about class membership to have broad and useful implications. This is the hardest part of ontology development. If you can find an existing ontology that has already undergone extensive use and refinement, it makes sense to adopt it.

It will be challenging to merge a collection of ontologies. Tool support will almost certainly be required to ensure a consistent KB.

sameClassAs, samePropertyAs

In order to tie together a set of component ontologies as part of a third it is frequently useful to be able to indicate that a particular concept in one ontology is equivalent to some concept in a second ontology. This capability must be used with care. It is fairly easy to create necessarily empty sets when combining a set of distributed definitions. If the combined ontologies are contradictory (all A's are B's vs. all A's are not B's) there will be no extension (sets of individuals and relations) that satisfies the resulting combination.

In the food ontology we want to link wine features in the descriptions of dining courses back to the wine ontology. One way to do this is by essentially redefining the concept as an element of owl:Thing and then linking it to the other resource.

<owl:Thing rdf:ID="#OffDry">
  <owl:sameClassAs rdf:resource="&vin;OffDry" />

Of course the example above is somewhat contrived, since we can always use &vin;OffDry anywhere we would #OffDry and get the same effect without redefinition. A more likely use would be in a case were I depend on two, independently developed ontologies, and note that they use the terms O1:foo and O2:bar to reference the same concept. sameClassAs could be used to collapse these together so that the entailments from the two ontologies become mutually supporting.

We have already seen that class expressions can be the target of subClassOf constructors. They can also be the target of sameClassAs [OWL/DL]. Again, this avoids the need to contrive names for every class expression and provides a powerful definitional capability based on satisfaction of a property.

<Region rdf:ID="TexasThings"> 
      <owl:onProperty rdf:resource="#locatedIn" />
      <owl:allValuesFrom rdf:resource="#TexasRegion" />
</Region>                                                ¬ 
TexasThings are exactly those things located in the Texas region.


This mechanism is similar to that for classes, but declares two individuals to be identical. An example would be:

<Wine rdf:ID="MikesFavoriteWine"> 
  <owl:sameIndividualAs rdf:resource="#StGenevieveTexasWhite" /> 
</Wine>                                                  ¬ 
This example does not have great utility. About all we learn from this is that Mike is not a connoisseur of wines. A more typical use of sameIndividualAs would be to equate individuals defined in different documents to one another, as part of unifying two ontologies.

This brings up an important point. OWL does not have a unique naming assumption. Just because two names are different does not mean they refer to different individuals. In the case above, we asserted identity between two distinct names. But it is just as possible for this sort of identity to be inferred.

Remember the implications that can be derived from a functional property. Given that hasMaker is functional, the following is not necessarily a conflict.

<owl:Thing rdf:about="#BancroftChardonnay">
  <hasMaker rdf:resource="#Bancroft" />
  <hasMaker rdf:resource="#Beringer" />
</owl:Thing>                                             ¬ 
It may simply mean that Bancroft = Beringer.


This mechanism provides the opposite effect from sameIndivdualAs.

<WineSugar rdf:ID="Dry" />

<WineSugar rdf:ID="Sweet">  
  <owl:differentIndividualFrom rdf:about="#Dry"/>  

<WineSugar rdf:ID="OffDry">
  <owl:differentIndividualFrom rdf:about="#Dry"/> 
  <owl:differentIndividualFrom rdf:about="#Sweet"/> 
This is one way to assert that these three values are mutually distinct. There will be cases where it is important to ensure such distinct identities. Without these assertions we could describe a wine that was both Dry and Sweet. We have stated that the hasSugar property applied to a wine has no more than one value. If we erred, and asserted that a wine was both Dry and Sweet, without the differentIndividualFrom statements above, this would imply that Dry and Sweet are identical. With the statements above, we would instead get a contradiction.

Complex Classes [OWL/DL]

OWL provides additional constructors with which to form classes. These constructors can be used to create so-called class expressions. OWL supports the basic set operations, namely intersection, union and complement. These are denoted unionOf, intersectionOf, and complementOf, respectively. Additionally, classes can be enumerated. Class extensions can be stated explicitly by means of the oneOf constructor. And it is possible to assert that class extensions must be disjoint.

Note that Class expressions can be nested without requiring the creation of names for every intermediate class. This allows the use of set operations to build up complex classes from anonymous classes or classes with value restrictions.

Set Operators
intersectionOf, unionOf, complementOf [OWL/DL]

Since OWL class extensions are sets, OWL provides the means to manipulate class extensions using basic set operators.

Intersection [OWL/DL]

The following examples demonstrate the use of the intersectionOf construct.

<owl:Class rdf:ID="WhiteWine">
    <owl:intersectionOf rdf:parseType="Collection">
      <owl:Class rdf:about="#Wine" />
        <owl:onProperty rdf:resource="#hasColor" />
        <owl:hasValue rdf:resource="#White" />
Classes constructed using the set operations are closed. This is an important capability. It permits us to state that WhiteWine is exactly the intersection of the class Wine and the set of things that are white in color. This means that if something is white and a wine, then it is an element of WhiteWine. Without such closed sets we can know that white wines are wines and white, but not vice-versa. This is an important tool for categorizing instances. (Note that 'rdf:parseType="Collection"' is a required syntactic element.)
<owl:Class rdf:about="#Burgundy">
    <owl:intersectionOf rdf:parseType="Collection">
      <owl:Class rdf:about="#Wine" />
        <owl:onProperty rdf:resource="#locatedIn" />
        <owl:hasValue rdf:resource="#BourgogneRegion" />

Here we define Burgundy to include exactly those wines that have at least one locatedIn relation to the Bourgogne Region. We could have declared a new class ThingsFromBourgogneRegion and used it as a class in the owl:intersectionOf construct. Since we do not have any other use for ThingsFromBourgogneRegion, the declaration above is shorter, clearer and doesn't require the creation of a contrived name.

<owl:Class rdf:ID="WhiteBurgundy">
    <owl:intersectionOf rdf:parseType="Collection">
      <owl:Class rdf:about="#Burgundy" />
      <owl:Class rdf:about="#WhiteWine" />

Finally the class WhiteBurgundy is exactly the intersection of white wines and Burgundies. Burgundies in turn are grown in the French region of Bourgogne and are dry wines. Accordingly all individual wines that meet these criteria are part of the class extension of WhiteBurgundy.

Union [OWL/DL]

The following example demonstrates the use of the unionOf construct. It is used exactly like the intersectionOf construct:

<owl:Class rdf:ID="Fruit">
    <owl:unionOf rdf:parseType="Collection">
      <owl:Class rdf:about="#SweetFruit" />
      <owl:Class rdf:about="#NonSweetFruit" />

The class Fruit includes both the extension of SweetFruit and the extension of NonSweetFruit.

Note how completely different this union type construct is from the following.

<owl:Class rdf:ID="Fruit">
  <rdfs:subClassOf rdf:resource="#SweetFruit" />
  <rdfs:subClassOf rdf:resource="#NonSweetFruit" />
</owl:Class>                                             ¬ 
This says that the elements of Fruit are in the intersection of sweet and non-sweet fruit, which we would expect to be the empty set.

Complement [OWL/DL]

The complementOf construct selects all individuals from the domain of discourse that do not belong to a certain class. Usually this refers to a vary large set of individuals:

  <owl:Class rdf:ID="ConsumableThing" />

  <owl:Class rdf:ID="NonConsumableThing">
    <owl:complementOf rdf:resource="#ConsumableThing" />

The class of NonConsumableThing includes as its members all individuals that do not belong to the extension of ConsumableThing. This set includes all Wines, Regions, etc. It is literally the set difference between owl:Thing and ConsumableThing. Therefore, a typical usage pattern for complementOf is in combination with other set operators:

<owl:Class rdf:ID="NonFrenchWine">
  <owl:intersectionOf rdf:parseType="Collection">
    <owl:Class rdf:about="#Wine"/>
        <owl:onProperty rdf:resource="#locatedIn" />
        <owl:hasValue rdf:resource="#FrenchRegion" />
</owl:Class>                                             ¬ 
This defines the class NonFrenchWine to be the intersection of Wine with the set of all things not located in France.

Enumerated Classes
oneOf [OWL/DL]

OWL provides the means to define a class via a direct enumeration its members. This is done using the oneOf constructor. Notably, this definition closes the class extension, so that no other individuals can be declared to belong to the class.

The following statement defines a class WineColor whose members are the individuals White, Rose, and Red.

<owl:Class rdf:ID="WineColor">
  <rdfs:subClassOf rdf:resource="#WineDescriptor"/>
  <owl:oneOf rdf:parseType="Collection">
    <owl:Thing rdf:about="#White"/>
    <owl:Thing rdf:about="#Rose"/>
    <owl:Thing rdf:about="#Red"/>

The first thing to understand here is that no other individuals can be a valid WineColor since the class has been defined by enumeration and is closed.

Each element of the oneOf construct must be a validly declared individual. An individual has to belong to some class. In the above example, each individual was referenced by name. We used owl:Thing as a simple cliché to introduce the reference. Alternatively, we could have referenced the elements of the set according to their specific type, WineColor, by:

<owl:Class rdf:ID="WineColor">
  <rdfs:subClassOf rdf:resource="#WineDescriptor"/>
  <owl:oneOf> rdf:parseType="Collection">
    <WineColor rdf:about="#White" />
    <WineColor rdf:about="#Rose" />
    <WineColor rdf:about="#Red" />

Other, more complex descriptions of individuals are also valid elements of the oneOf construct, for example:

<WineColor rdf:about="#White">

Disjoint Classes
disjointWith [OWL/DL]

The disjointness of a set of classes can be expressed using the owl:disjointWith constructor. It guarantees that an individual that is a member of one class cannot simultaneously be an element of a specified other class.

<owl:Class rdf:ID="NonSweetFruit"> 
  <owl:disjointWith rdf:resource="#SweetFruit"/> 

<owl:Class rdf:ID="Pasta">
  <rdfs:subClassOf rdf:resource="#EdibleThing"/>
  <owl:disjointWith rdf:resource="#Meat"/>
  <owl:disjointWith rdf:resource="#Fowl"/>
  <owl:disjointWith rdf:resource="#Seafood"/>
  <owl:disjointWith rdf:resource="#Dessert"/>
  <owl:disjointWith rdf:resource="#Fruit"/>

The first statement ensures that any fruit is either a NonSweetFruit or a SweetFruit. An inconsistency would arise if it were both. (We in fact defined Fruit to be the union of these two classes.)

The Pasta example demonstrates multiple disjoint classes. Note that this only asserts that Pasta is disjoint from all of these other classes. It does not assert, for example, that Meat and Fruit are disjoint. In order to assert that a set of classes are mutually disjoint, there must be an owl:disjointWith assertion for every pair.

A common requirement is to define as a class as the union of a set of mutually disjoint subclasses.

<owl:Class rdf:ID="SweetFruit">
  <rdfs:subClassOf rdf:resource="#EdibleThing" />

<owl:Class rdf:ID="NonSweetFruit">
  <rdfs:subClassOf rdf:resource="#EdibleThing" />
  <owl:disjointWith rdf:resource="#SweetFruit" />

<owl:Class rdf:ID="Fruit">
    <owl:unionOf rdf:parseType="Collection">
      <owl:Class rdf:about="#SweetFruit" />
      <owl:Class rdf:about="#NonSweetFruit" />
Here we define Fruit to be exactly the union of SweetFruit and NonSweetFruit. And we know that these subclasses exactly partition Fruit into two distinct subclasses because they are disjoint. As the number of mutually disjoint classes grows, the number of disjointness assertions grows proportionally to n2. However, in the use cases we have seen, n is typically small.

Usage Examples

Once an initial domain ontology is available, a large number of applications can be developed that exploit the ontology. In this section, we describe some sample uses in the domain of wines.

Wine Portal

A number of sites exist today that call themselves wine portals. Google for example, provides 152,000 matches for the query "wine portal". One of the top matches, a site called "Wine-Portal.com", provides access to a number of sites. Many of sites claiming to be wine portals are mostly informational sites. For example, wine-portal.com's first featured site, called 'cork cuisine' (www.corkcuisine.com/), provides information about matching wines and foods, wines as gifts, etc. Another site billed as "the Internet Wine Portal" ( www.cyberbacchus.com/) provides a nice organization of a substantial amount of wine information on numerous topics.

Perusing any of the topic areas, one finds a collection of pages containing information and sometimes services related to the topic. For example, 'accessories and gifts' contains information about what to look for when buying particular wine items and also contains a significant number of online retailers. Another top level area called 'shopping' has a subarea called 'wine shopping' from which a user can find online (or 'street shopping') stores (categorized by country). These two sites are just two of the many examples today and are representative of the general notion of a wine portal providing a collection of information and services connected to a particular topic area.

When looking at these sites in some detail, it is not clear how much they depend on ontologies today. For example, viewing the source for the html does not reveal evidence of ontological usage. However, it is clear that the sites could exploit ontologies had some wine ontologies been available.

One simple use of ontologies in portal sites is for organization and browsing. The listing of categories above could be generated from the top few levels of wine related classes. Queries could exploit wine ontologies to retrieve wine relevant information. If one did a search for a term contained in the ontology, the query could be expanded with subclass information in order to find more relevant answers. Portals could be made to automatically update themselves with (candidate) information in topic areas. With very powerful reasoning capabilities they could even identify likely wine sales sites and negotiate to include them as part of the portal.

Wine Agent

We have started a wine agent for expository purposes. In our initial design, the wine agent's goal is to recommend wines to accompany meal courses. This application exploits the ontology used as the basis of this guide. This wine ontology is available in the DAML ontology library and is entitled wines.

A personalized wine agent can provide a number of services for a human.
The agent may be used to recommend wines given a set of constraints (such as a meal being served), the agent may find information about a particular wine or a particular class of wines, it may look for appropriate accessories for a wine (such as a particular kind of glass suited for that wine varietal, etc.).

Below, we describe an example in a simple prototype system that is being written as a student project.

Consider the following scenario:

Someone is planning a dinner party and at least one of the guests is wine knowledgeable. The host would like to serve wine that is well matched to the course(s) on the menu. The host would also like to appear knowledgeable about the wines served at the event. The host would also like to have appropriate accessories at the dinner. The host may have decided to serve a special tomato based pasta sauce with fresh pasta as the main course.

In order to serve wines appropriate to the meal, the host needs information concerning wine and food pairings. In order to appear knowledgeable about wines, the host would benefit from having access to wine information relevant to the event. In order to have appropriate wine accessories, the host would need to have information about what accessories are relevant to the situation (and are within the host's price range).

With a background wine ontology, given a description of a meal, a wine agent can suggest the type of wine to serve with the meal. The wine agent may suggest a zinfandel as the varietal of choice for the meal. Additionally, given a background ontology, the wine agent may suggest a particular zinfandel, possibly Marietta Zinfandel. Given the information that the wine should be a zinfandel, a wine agent may look for a place to acquire either a selection of zinfandels or it may look for a particular zinfandel wine, such as Marietta. Given a background ontology containing appropriate sources for wine purchases (possibly filtered by the location of the host and the location of the wine seller), the wine agent could go to a site such as wine.com and do a search for zinfandels returning a listing of zinfandels for sale on that site. The wine agent could attempt to find Marietta Zinfandel either from the winery itself or from other resellers. It could, for example, find (by a search on Google or a structured search of selected web sites) that winelibrary.com has a sale on Marietta Zinfandel 1999 vintage for a discounted price of $13.99. The wine agent could use additional filtering information such as price ranges provided either by the consumer or as suggestions based on varietal.

The wine agent may now attempt to provide information concerning zinfandel in general or Marietta Zinfandel in particular. It could use a background ontology of wine sites to find information about particular wines. For example, the winery description of its 1999 Zinfandel may be of use. Additionally reviews from respected sources such as the Wine Spectator may be of use. If no review of Marietta Zinfandel is available on a favorite wine review site, it may be useful to look for related information such as reviews on zinfandels from the same region, in this case Zinfandels from Sonoma County, California.

General background information may also be of use. The host may also want to do some reading and may be interested in books on wine in general or zinfandels in particular. For example, the host may be interested in the books that Amazon.com has for sale on zinfandel. The host may also be interested in information concerning wines from the same region, and thus may be interested in Sonoma Zinfandels. A wine agent may have typical background information available that is related to its main knowledge areas. For example, this wine agent is concerned with matching foods and wines, so it may have both free and purchasable information on this topic such as the Wine Spectator's article on matching food and wine.

The dinner host may also want to acquire appropriate wine accessories prior to the event. Wine is served in wine glasses and different wine varietals are best served in different kinds of glasses. For example, if the host has chosen a meal course for which a zinfandel is appropriate, the host may want to know that Riedel is a well-known manufacturer of wine glassware. The host may also want to be linked to the Wine Enthusiast (a well respected supplier of wine merchandise) and be told that the Wine Enthusiast has Riedel's Vinum Zinfandel glass for sale as a set of 4 for $63.95 (with a discount to $59.95 if you buy two sets of 4 glasses). The host may also be interested to know that Amazon.com has Reidel's Sommelier Zinfandel single stem glass available for $49.99 (and claims a list price of $65.00). Amazon also has the same Vinum glass for sale in sets of 6 (instead of 4 on the wine enthusiast) for $79.99 (and claims a list price of $119.40). A wine agent could provide a comparison listing of glassware that is matched to the meal (i.e., is appropriate to be used to serve zinfandel) and then is compared by price or other criteria chosen from a list of properties in the ontology.

The dinner host may want to consider other wine accessories. From the ontology, we know that corkscrews are wine accessories. The background ontology may encode subclasses of corkscrews or such information could be found from relevant wine sites as well. The Wine Enthusiast has a set of corkscrews they recommend (with descriptions of the types and price ranges). They also distinguish corkscrews by type (level, waiter, stationary, twist, and pump) and the dinner host may want to get information about those styles.

The wine agent may be taken to many levels of sophistication depending upon background ontology knowledge of the domain and information and services sites. In this example, we only exploited information concerning wines, varietal type, food and wine combinations, some wine accessories and their related properties. We could of course expand this to include more information and more constraints by the customer.

An evolving example of this wine agent will soon be available.


This section describes the status of this document at the time of its publication. Other documents may supersede this document.

This document is a working document for the use by W3C Members and other interested parties. It may be updated, replaced or made obsolete by other documents at any time.

This document has been produced as part of the W3C Semantic Web Activity, following the procedures set out for the W3C Process. The document has been compiled by the Web Ontology Working Group. The goals of the Web Ontology working group are discussed in the Web Ontology Working Group charter.

A list of current W3C Recommendations and other technical documents can be found at http://www.w3.org/TR/.


Related OWL

Requirements for a Web Ontology Language. W3C Working Draft, 08 July 2002.

Feature List
Feature Synopsis for OWL Lite and OWL. Deborah L. McGuinness and Frank van Harmelen. W3C Working Draft, 29 Jul 2002.

OWL Web Ontology Language 1.0 Reference. Mike Dean, Dan Connolly, Frank van Harmelen, James Hendler, Ian Horrocks, Deborah L. McGuinness, Peter F. Patel-Schneider, and Lynn Andrea Stein. W3C Working Draft, 29 July 2002.

XML Syntax
[@@ ref]

UML Syntax
[@@ ref]

Abstract Syntax
OWL Web Ontology Language 1.0 Abstract Syntax. Peter F. Patel-Schneider, Ian Horrocks, and Frank van Harmelen. W3C Working Draft 29 July 2002.

Formal Model
Model-Theoretic Semantics for OWL, Peter F. Patel-Schneider, 15 August 2002

Sample Ontologies and Applications

Daml.org Wine Ontology

Wine Ontology Tutorial
Ontology Development 101: A Guide to Creating Your First Ontology Natalya Fridman Noy and Deborah L. McGuinness. Stanford Knowledge Systems Laboratory Technical Report KSL-01-05 and Stanford Medical Informatics Technical Report SMI-2001-0880, March 2001.

Wine Ontology in CLASSIC
Living with CLASSIC: When and How to Use a KL-ONE-Like Language , Ronald J. Brachman, Deborah L. McGuinness , Peter F. Patel-Schneider , Lori Alperin Resnick , and Alex Borgida. in John Sowa, ed., Principles of Semantic Networks: Explorations in the representation of knowledge, Morgan-Kaufmann: San Mateo, California, 1991, pages 401--456.
Industrial Strength Ontology Management. Aseem Das, Wei Wu, and Deborah L. McGuinness. Stanford Knowledge Systems Laboratory Technical Report KSL-01-09 2001. In the Proceedings of the International Semantic Web Working Symposium. Stanford, CA, July 2001

Related Ontology Research

DAML+OIL W3C Submission. Includes reference description, both model theoretic and axiomatic semantics, annotated walkthrough and examples.

Annotated DAML+OIL Ontology Markup. Frank van Harmelen, Peter F. Patel-Schneider and Ian Horrocks (eds), March 2001.

Daml Home Page

DAML-ONT initial release
DAML-ONT Partial DAML-ONT axiomatization in KIF

OIL Home Page

Ontoknowledge Home Page

Simple HTML Ontology Extensions (SHOE) home page. University of Maryland.

The Ontobroker home page. Institute AIFB, University of Karlsruhe.

XOL: An XML-Based Ontology Exchange Language. Karp, Peter D., Chaudhri, Vinay K. and Thomere, Jerome F. Technical Report 559. AI Center, SRI International, 333 Ravenswood Ave., Menlo Park, CA 94025, Jul 1999.

The Conceptual Knowledge Framework: OML/CKML. Kent, Robert. KAW'99, Twelfth Workshop on Knowledge Acquisition, Modeling and Management, Voyager Inn, Banff, Alberta, Canada, October 16-21, 1999.

OKBC: A programmatic foundation for knowledge base interoperability, V. K. Chaudhri, A. Farquhar, R. Fikes, P. D. Karp, and J. P. Rice. In Proceedings of the 15th National Conference on Artificial Intelligence (AAAI-98), pages 600?607. AAAI Press,1998.

Meta Content Framework Using XML. Guha, R.V. and Bray, Tim. Netscape Communications, 6 June 1997.

Related W3C Standards

Resource Description Framework (RDF) Model and Syntax Specification. World Wide Web Consortium Recommendation, 1999 Lassila, Swick [eds].

RDF: Resource Description Framework. Background at W3C.

RDF Vocabulary Description Language 1.0: RDF Schema. W3C Working Draft 30 April 2002.

Extensible Markup Language (XML) 1.0

XML Tutorial 1: Well-Formed XML Documents. Bonnie SooHoo Aug 4, 2000 in webreview.com.

Extensible Markup Language (XML). More background at W3C.

XML Schema
W3C XML Schema

Uniform Resource Identifiers (URI): Generic Syntax. IETF Draft Standard August 1998 (RFC 2396) T. Berners-Lee, R. Fielding, L. Masinter

A Beginner's Guide to URLs. The classic intro to URLs, by the NCSA Mosaic team.

Web Naming and Addressing Overview (URIs, URLs, ...). More background at W3C.

XML Namespaces
Namespaces in XML. W3C Recommendation Jan 1999.

XML Namespaces by Example. Tim Bray Jan. 19, 1999 in XML.com.

Namespace Myths Exploded. Ronald Bourret. March 8, 2000 in XML.com.

XML Base
XML Base. W3C Recommendation


Description Logics
Franz Baader, Diego Calvanese, Deborah L. McGuinness, Daniele Nardi, and Peter F. Patel-Schneider, editors, The Description Logic Handbook: Theory, Implementation and Application. Cambridge University Press, 2002.

Description Logic Web Page

Knowledge Representation and Ontology
Dublin Core Metadata

Expressing Dublin Core in RDF/XML

KR Home Page

Part Whole Relations
Winston, Chaffin & Herrmann. A Taxonomy of Part-Whole Relations. Cognitive Science, 11:417-444, 1987.

Controlled Vocabularies
Mind your phraseology! Using controlled vocabularies to improve findability. DigitalWeb Magazine, Christina Wodtke

Appendix A: XML + RDF Basics

This appendix provides links to introductions to the standards that OWL depends on.

To fully understand the OWL syntax and semantics you should be familiar with the basics of the related W3C and IETF standards listed below. A minimal guide to XML and RDF is provided by the first two links below.

Appendix B: History

The Resource Description Framework (RDF) was the first language specified by the W3C for representing semantic information about arbitrary resources. RDF Schema (RDFS) is a W3C candidate recommendation for an extension to RDF to describe RDF vocabularies. RDFS can be used to create ontologies, but it is purposefully lightweight, with less expressive power than OWL.

Like OWL, RDFS includes classes and properties, as well as range and domain constraints on properties. It provides inheritance hierarchies for both classes and properties. Upon its release users began requesting additional features, including data types, enumerations and the ability to define properties more rigorously.

Other efforts in the research community were already examining exactly these sorts of features. For those who wish to delve more deeply into this background, a partial list of projects and languages includes:

Instead of continuing with separate ontology languages for the Semantic Web, a group of researchers, including many of the main participants in both the OIL and DAML-ONT efforts, got together in the Joint US/EU ad hoc Agent Markup Language Committee to create a new web ontology language. This language DAML+OIL built on both OIL and DAML-ONT, was submitted to the W3C as a proposed basis for OWL, and was subsequently selected as the starting point for OWL.

In addition to ontology languages, various taxonomies and existing ontologies are already in use commercially. In e-Commerce sites they facilitate machine-based communication between buyer and seller, enable vertical integration of markets and allow descriptions to be reused in different marketplaces. Examples of sites that are actually making commercial use ontologies include:

Various medical or drug-related ontologies have been developed to help manage the overwhelming mass of current medical and biochemical research data that can be difficult to tie together into a cohesive whole. One major resource is the Gene Ontology Consortium which is defining ontologies for

That site also has pointers to ontologies for

There exist large taxonomies in use today that would be ripe for extension into the OWL space. For example, the North American Industry Classification System (NAICS) defines a hierarchy of over 1900 elements that identify industry types. NAICS is also tied to the International Standard Industrial Classification System (ISIC, Revision 3), developed and maintained by the United Nations.

Appendix C: An Alternate Region Ontology

This example was developed by Guus Schrieber [ W3C WG Archive] and presents a more elaborate wine region ontology.

For wine the "production area" is an important feature. There is enormous variation between wine types with respect to the grain size of the production area, ranging from a complete country to a particular vineyard. One could distinguish four types of production area's:

  1. country: e.g., France, Italy
  2. region: e.g., Bordeaux, Medoc, Tuscany,
  3. town: e.g., Margaux, Montalcino, Montepulciano
  4. vineyard, e.g., Chateau Margaux, Avignonesi

In addition, we need to model part-of relationships between the various sorts of production areas:

We would like to be able to derive from our wine KB that a wine from Chateau Margaux is a French wine and that Avignonesi is a Tuscan wine.

MODELING DECISION: at this point we decided to drop the "town" subclass and treat towns as regions. This simplifies the model and is consistent with the fact that a "town" as a wine production area typically stands for an area surrounding the town, which can be larger or smaller that the actual town area. For example, the production area "Montalcino" is in fact a subregion of Tuscany surrounding the village of Montalcino.

This leads to the following model:


<owl:Class rdf:ID="&vin;ProductionArea"/ >
<owl:Class rdf:ID="&vin;Country:">
    <rdfs:subClassOf rdf:resource="&vin;ProductionArea"/>
<owl:Class rdf:ID="&vin;Region:"> 
    <rdfs:subClassOf rdf:resource="&vin;ProductionArea"/> 
<owl:Class rdf:ID="&vin;Vineyard:"> 
    <rdfs:subClassOf rdf:resource="&vin;ProductionArea"/> 

vin:ProductionArea rdf:type rdfs:Class.
vin:Country rdfs:subClassOf vin:ProductionArea.
vin:Region rdfs:subClassOf vin:ProductionArea.
vin:Vineyard rdfs:subClassOf vin:ProductionArea.


<owl:ObjectProperty rdf:ID="&vin;hasSubArea">   

<owl:ObjectProperty rdf:ID="&vin;subAreaOf">   
    <owl:inverseOf rdf:resource="&vin;hasSubArea"/> 

<owl:ObjectProperty rdf:ID="&vin;hasRegion">   
    <rdfs:subPropertyOf rdf:resource="&vin;hasSubArea"/> 
    <owl:allValuesFrom rdf:resource="&vin;Region"/>

<owl:ObjectProperty rdf:ID="&vin;regionOf">   
    <rdfs:inverseOf rdf:resource="&vin;hasRegion"/> 
    <owl:allValuesFrom rdf:resource="&vin;Country"/>

<owl:ObjectProperty rdf:ID="&vin;hasSubRegion">   
    <rdfs:subPropertyOf rdf:resource="&vin;hasSubArea"/> 
    <owl:allValuesFrom rdf:resource="&vin;Region"/>

<owl:ObjectProperty rdf:ID="&vin;subRegionOf">   
    <rdfs:inverseOf rdf:resource="&vin;hasSubRegion"/> 
    <owl:allValuesFrom rdf:resource="&vin;Region"/>

<owl:ObjectProperty rdf:ID="&vin;hasVineyard">   
    <rdfs:subPropertyOf rdf:resource="&vin;hasSubArea"/> 
    <owl:allValuesFrom rdf:resource="&vin;Vinyard"/>

<owl:ObjectProperty rdf:ID="&vin;vineyardRegion">   
    <rdfs:inverseOf rdf:resource="&vin;hasVineyard"/> 
    <owl:allValuesFrom rdf:resource="&vin;Region"/>
vin:hasSubArea rdf:type rdfs:Property.  
vin:hasSubArea rdf:type owl:TransitiveProperty. 
vin:subAreaOf owl:inverseOf vin:hasSubArea. 

vin:hasRegion rdfs:subPropertyOf vin:hasSubArea.
vin:hasRegion owl:allValuesFrom vin:Region.
vin:regionOf owl:inverseOf vin:hasRegion.
vin:regionOf owl:allValuesFrom vin:Country.
vin:regionOf owl:cardinality 1.

vin:hasSubRegion rdfs:subPropertyOf vin:hasSubArea.
vin:hasSubRegion owl:allValuesFrom vin:Region.
vin:subRegionOf owl:inverseOf vin:hasSubRegion.
vin:subRegionOf owl:allValuesFrom vin:Region.
vin:subRegionOf owl:cardinality 1.

vin:hasVineyard rdfs:subPropertyOf vin:hasSubArea.
vin:hasVineyard owl:allValuesFrom vin:Vineyard.
vin:vineyardRegion owl:inverseOf vin:hasVineyard..
vin:vineyardRegion owl:allValuesFrom vin:Region.
vin:vineyardRegion owl:cardinality 1.

ONTOLOGICAL NOTE: the part-whole relationship described here is well-known in the formal-ontology literature. The typology of part-whole relations by Winston et al. characterizes this as a "place-area" relation. Various authors have criticized and amended this typology. If the formal-ontology community would at some point be able to make a part-whole typology available in OWL, the properties in this wine ontology could be linked to it.

UML NOTE: Below is a UML class diagram for this example. The modeling decisions for this diagram will be discussed in the forthcoming document on the UML presentation syntax. For the moment, note the use of the UML "composition" construct (the darkened diamond), which caries some of the semantics of the place-area relation.