3/6/08. Introduction to Ontology Web Language (OWL) Graphs. Representing knowledge. Graphs as knowledge 2. Graphs as knowledge 1

Transcription

1 University of Dublin Trinity College Stack Architecture for Semantic Web Introduction to Ontology Web Language (OWL) Dr. Owen Conlan Intro to OWL Declan O Sullivan 2 Representing knowledge There are a number of options As objects, using the well-accepted techniques of object-oriented analysis and design to capture a model As clauses, going back to the early days of AI and Lisp As XML, using the industry-standard structured mark-up language As graphs, making use of the things we know about graph theory As some combination of these We are looking for: extensibility, ease of use, ease of querying Oriel 3.16 Simon office Person1234 Graphs We can use the nodes of a graph for facts and the arcs as (binary) relationships between them Arcs are typically called predicates or relationships in this view The set of arcs intersecting a node tells us the information we know about that fact or entity Which would you choose? Intro to OWL Simon Declan Dobson O Sullivan 3 Intro to OWL Simon Declan Dobson O Sullivan 4 Graphs as knowledge 1 Graphs as knowledge 2 How do we use graphs to represent knowledge? Dobson A key from Software which to hang Simon engineering the different firstname facts teaches Person1234 teaches office Contextual extension systems takes Oriel 3.16 x3681 takes Student 2 phone Student 1 firstname Joe Bloggs Things to note Scaling the same graph can represent a load of different knowledge simultaneously Agreement need to know what the various predicates mean Structure you need to know what nodes are related by a predicate Plurality the same relationship may appear several times Symmetry the same predicates can be used for common information, despite minor changes Asymmetry relationships are inherently directed, which sometimes makes things awkward and this can get very tricky and this can be difficult to keep straight For example both lecturers and students have names So a knowledge (context) graph is inherently directed Intro to OWL Simon Declan Dobson O Sullivan 5 Intro to OWL Simon Declan Dobson O Sullivan 6 1

2 Two ways to view a graph As nodes and arcs Nodes store facts, edges store relationships between them Simon Person1234 As triples A three-place relationship of subject, predicate, object The node and edge structure is induced by the triples each triple defines an edge, the different subjects/objects are the population of nodes, one node per individual string Person1234 Simon Subject Predicate Resource Description Framework (RDF) RDF is a W3C recommendation that enables encoding, exchange and reuse of structured metadata in XML Triples of assertions can be expressed using XML tags Concept RDF statement Value relationship Concept E.g. Cabernet Sauvignon grape, is a type of, Wine grape <rdf:description rdf:about="cabernet Sauvignon grape > <rdf:type rdf:resource= #Wine grape" /> </rdf:description> Object Each resource can be assigned a different Universal Resource Identifier (URI) Thus different meanings for the same term can be assigned different URIs Reference: RDF Primer. W3C draft technical note, 2002 Intro to OWL Simon Declan Dobson O Sullivan 7 Intro to OWL Declan O Sullivan 8 URI = Uniform Resource Identifier URIs "The generic set of all names/addresses that are short strings that refer to resources" URLs (Uniform Resource Locators) are a particular type of URI, used for resources that can be accessed on the WWW (e.g., web pages) In RDF, URIs typically look like normal URLs, often with fragment identifiers to point at specific parts of a document: A cluster of facts Given a common subject we can build a cluster of facts using nested predicate elements <rdf:rdf xmlns:rdf= xmlns:s= > <rdf:description about= > <s:about>distributed Systems Group</s:about> <s:author>s. Punter</s:author> <s:phone> </s:phone> </rdf:description> </rdf:rdf> As long as we agree what the predicates mean, we can use whichever we want Each of these gives rise to a triple with the same subject (inherited from the containing Description element) Intro to OWL Declan O Sullivan 9 Intro to OWL Simon Declan Dobson O Sullivan 10 Structuring the knowledge RDF provides a way of building graphs from triples, but doesn t constrain the graph too much Nothing stops an application from giving a place a, for example, although this is probably nonsense The problem is that RDF is an untyped mechanism for building graphs No knowledge of which triples are allowed, or what thing must be the subject/object of an arc This is a problem in two distinct ways In interpretation different people may interpret the predicates subtly differently and use them between values you can t handle In scaling hard for an application to get it right RDF and RDFS RDF stands for Resource Description Framework It is a W3C candidate recommendation ( RDF is graphical formalism ( + XML syntax + semantics) for representing metadata for describing the semantics of information in a machineaccessible way RDFS extends RDF with schema vocabulary, e.g.: Class, Property type, subclassof, subpropertyof range, domain Intro to OWL Simon Declan Dobson O Sullivan 11 Intro to OWL Declan O Sullivan 12 2

3 RDFS Examples RDF Schema terms (just a few examples): Class Property type subclassof range domain These terms are the RDF Schema building blocks (constructors) used to create vocabularies: <Person,type,Class> <hascolleague,type,property> <Professor,subClassOf,Person> <Carole,type,Professor> <hascolleague,range,person> <hascolleague,domain,person> Intro to OWL Declan O Sullivan 13 RDF/RDFS Liberality No distinction between classes and instances (individuals) <Species,type,Class> <Lion,type,Species> <Leo,type,Lion> Properties can themselves have properties <hasdaughter,subpropertyof,haschild> <hasdaughter,type,familyproperty> No distinction between language constructors and ontology vocabulary, so constructors can be applied to themselves/each other <type,range,class> <Property,type,Class> <type,subpropertyof,subclassof> Intro to OWL Declan O Sullivan 14 Problems with RDFS RDFS too weak to describe resources in sufficient detail No localised range and domain constraints Can t say that the range of haschild is person when applied to persons and elephant when applied to elephants No existence/cardinality constraints Can t say that all instances of person have a mother that is also a person, or that persons have exactly 2 parents No transitive, inverse or symmetrical properties Can t say that ispartof is a transitive property, that haspart is the inverse of ispartof or that touches is symmetrical Difficult to provide reasoning support No native reasoners for non-standard semantics Web Ontology Language Requirements Desirable features identified for Web Ontology Language: Extends existing Web standards Such as XML, RDF, RDFS Easy to understand and use Should be based on familiar KR idioms Formally specified Of adequate expressive power Possible to provide automated reasoning support Intro to OWL Declan O Sullivan 15 Intro to OWL Declan O Sullivan 16 From RDF to OWL Two languages developed to satisfy above requirements OIL: developed by group of (largely) European researchers (several from EU OntoKnowledge project) DAML-ONT: developed by group of (largely) US researchers (in DARPA DAML programme) Efforts merged to produce DAML+OIL Development was carried out by Joint EU/US Committee on Agent Markup Languages Extends ( DL subset of) RDF DAML+OIL submitted to W3C as basis for standardisation Web-Ontology (WebOnt) Working Group formed WebOnt group developed OWL language based on DAML+OIL OWL language now a W3C Candidate Recommendation Will soon become Proposed Recommendation Intro to OWL Declan O Sullivan 17 OWL Language Three species of OWL OWL full is union of OWL syntax and RDF OWL DL restricted to FOL fragment OWL Lite is easier to implement subset of OWL DL OWL DL Benefits from many years of DL research Well defined semantics Formal properties well understood (complexity, decidability) Known reasoning algorithms Implemented systems (highly optimised) Intro to OWL Declan O Sullivan 18 3

4 Ontology Representation Language Approach: Ontology Web Language (OWL) Being defined by W3C, <owl:class rdf:id="wine"> <rdfs:subclassof rdf:resource="#potableliquid"/> <owl:onproperty rdf:resource="#madefromgrape"/> <owl:mincardinality>1</owl:mincardinality> </rdfs:subclassof> <owl:onproperty rdf:resource="#locatedin"/> <owl:mincardinality>1</owl:mincardinality> </rdfs:subclassof>... Intro to OWL Declan O Sullivan 19 XX Potable Liquid Wine Madefrom Grape Locatedin YY subclass Define the term Room Example: Defining terms and a subclass relationship <owl:class rdf:id="room /> Define term Restroom and state that a Restroom is a type of Room <owl:class rdf:id="restroom"> <rdfs:subclassof rdf:resource="#room"/> Note: owl:thing is a predefined OWL Class and is the root of all classes Intro to OWL Declan O Sullivan 20 Defining Classes OWL provides several other mechanisms for defining classes equivalentclass allows you to state that two classes are synonymous disjointwith allows you to state that an instance of this class cannot be an instance of another E.g. Man and Woman could be stated as disjoint classes unionof allows you specify that a class contains things that are from more than one class E.g. Restroom could be defined as a union of MensRoom and LadiesRoom intersectionof allows you to specify that a class contains things that are both in one and the other complementof allows you specify that a class contains things that are not other things E.g. Children are not SeniorCitizens Example: equivalentclass and unionof <owl:class rdf:id="atomicplaceinbuilding"> <owl:class rdf:about="#atomicplace"/> </rdfs:subclassof> <owl:equivalentclass> <owl:class> <owl:unionof rdf:parsetype="collection"> <owl:class rdf:about="#room"/> <owl:class rdf:about="#hallway"/> <owl:class rdf:about="#stairway"/> <owl:class rdf:about="#otherplaceinbuilding"/> </owl:unionof> </owl:equivalentclass> Intro to OWL Declan O Sullivan 21 Intro to OWL Declan O Sullivan 22 Example disjointwith <owl:class rdf:id="stairway"> <rdfs:subclassof rdf:resource="#atomicplaceinbuilding"/> <owl:disjointwith rdf:resource="#room"/> <owl:disjointwith rdf:resource="#hallway"/> Associating a Property with a Class Associate the property (defined elsewhere) with the Class by using combination of <owl:onproperty> and defining local restrictions on that property using <owl:class rdf:id="restroom"> <rdfs:subclassof rdf:resource="#room" /> <owl:restriction owl:cardinality="1"> <owl:onproperty> <owl:objectproperty rdf:about="#accessrestrictedtogender" /> </owl:onproperty> <owl:allvaluesfrom> <owl:class rdf:about="#gender" /> </owl:allvaluesfrom> </owl:subclassof> </owl:class> Intro to OWL Declan O Sullivan 23 Intro to OWL Declan O Sullivan 24 4

5 Describe the following knowledge using OWL Academics are employees who work in a University Department Candidate OWL elements to use <owl:class> <owl:onproperty> <owl:allvaluesfrom> <rdf:about> <rdf:id> <owl:class rdf:id= Academic"> Over to you <rdfs:subclassof rdf:resource="#employee" /> <owl:onproperty> <owl:objectproperty rdf:about="#universitydept" /> </owl:onproperty> <owl:allvaluesfrom> <owl:class rdf:about="#departments" /> </owl:allvaluesfrom> </rdfs:subclassof> </owl:class> Intro to OWL Declan O Sullivan 25 Defining Properties In RDF Schema the rdf:property is used both to Relate one Resource to another Resource For example, a accessrestrictedtogender property can relate a Restroom to a Gender Relate a resource to a rdfs:literal or datatype For example, a latitude property relates a Room to a xsd:string type OWL provides different statements for two cases owl:objectproperty is used to relate a resource to another What range of values <owl:objectproperty rdf:id="accessrestrictedtogender"> `what <rdfs:range rdf:resource="#gender"/> classes is this <rdfs:domain rdf:resource="#restroom"/> property associated with owl:datatypeproperty is used to relate a resource to a rdfs:literal or XML schema data type <owl:datatypeproperty rdf:id="latitude"> <rdfs:range rdf:resource=" <rdfs:domain Intro to OWL rdf:resource="#place"/> Declan O Sullivan 26 Characterising Properties OWL allows use of three of the RDFS statements <rdfs:range> used to indicate the possible value types for a property. <rdfs:domain> use to associate a property with a class. <rdfs:subpropertyof> use this to specialize a property. In addition OWL allows you state that a property is Symmetric <rdf:type rdf:resource=" If a property is symmetric, then if the pair (x.y) is an instance then it can be deduced that (y,x) is also an instance. Example, property adjacentroom property of Room or friend property of person Transitive <rdf:type rdf:resource=" If a property is transitive, then if the pair (x.y) is an instance of P and (y,z) is an instance of P then it can be deduced that (x,z) is also an instance. Example, property isspatiallysubsumedby property <owl:objectproperty rdf:id="isspatiallysubsumedby" rdf:type=" <rdfs:domain rdf:resource=" <rdfs:range rdf:resource=" Intro to OWL Declan O Sullivan 27 Intro to OWL Declan O Sullivan 28 In addition OWL allows you state that a property is Functional If a property is functional, then it has no more than one value for each individual, that is equivalent to owl:maxcardinality restriction of 1 Example, property Address of a Room <owl:datatypeproperty rdf:id= Address" rdf:type=" <rdfs:domain rdf:resource="#room"/> </owl:datatypeproperty> In addition OWL allows you state that a property is InverseOf One property may be stated to be the inverse of another property. It can be deduced that if the property P1 is stated to be the inverse of P2, then if X is related to Y via P2 then Y is related to X via P1 Example, property spatiallysubsumes <owl:objectproperty rdf:id="spatiallysubsumes"> <owl:inverseof> <owl:objectproperty rdf:about="#isspatiallysubsumedby"/> </owl:inverseof> InverseFunctionalProperty The inverse of the property has most one value for each individual Example, property coordinates property of a Room could be declared as inverse functional, as the inverse theroomat property can only have one value <rdf:type rdf:resource=" Intro to OWL Declan O Sullivan 29 Intro to OWL Declan O Sullivan 30 5

6 <?xml version 1.0?> <Room rdf:id= LargeConferenceRoom > <address rdf:resource= G.02 /> Summary Example <spatiallysubsumedby rdf:resource= O Reilly Institute /> <adjacentroom rdf:resource= SmallConferenceRoom /> <coordinates rdf:resource=44,55 /> </Room> Given the preceding definitions it can be inferred automatically: 1. The O Reilly Institute spatially subsumes the Large Conf Room (since spatiallysubsumedby is an inverse property) 2. The Small Conference Room is adjacent to Large Conference Room (since adjacentroom is symmetric) 3. Only the Large Conference Room can be found at coordinates 44,55 (since coordinates is inverse functional) 4. The Large Conference Room has only one address (since address is functional) Intro to OWL Declan O Sullivan 31 Forward Chaining/Data Driven Reaction to new information is to examine a specific rule and act accordingly The temperature has changed => Check if the heater should be changed Inference Engine/Reasoner Software tool that aims to emulate the human capability to deduce new knowledge from already specified knowledge by reasoning e.g. RACER, Jena, etc Backward Chaining/Goal Driven Applies to situations where you are making a diagnostics. It presumes that we are aware of a given situation but we do not know what caused it Therefore we have to ask questions and/or perform some tests to determine the cause. It is too hot => Why? Is the heater on? Yes Turn it off Intro to OWL Declan O Sullivan 32 Going beyond basic OWL inferencing Semantic Web Rule Language SWRL Submitted to W3C (May 2004) OWL axioms are extended to include rules rules are of the form of an implication between an antecedent (body) and consequent (head); The intended meaning can be read as: whenever the conditions specified in the antecedent hold, then the conditions specified in the consequent must also hold. antecedent => consequent antecedent (body) and consequent (head) consist of zero or more atoms: Multiple atoms are treated as a conjunction An empty antecedent is treated as trivially true (i.e. satisfied by every interpretation), so the consequent must also be satisfied by every interpretation; an empty consequent is treated as trivially false (i.e., not satisfied by any interpretation), so the antecedent must also not be satisfied by any interpretation. Intro to OWL Declan O Sullivan 33 Rules Form Atoms can take the forms: C(x), P(x,y), sameas(x,y), differentfrom(x,y), builtin(r,x,...) C is an OWL description or data range; P is an OWL property; r is a built-in relation; (can write in normal functional form if function) x and y are either: variables (normally preceded by? Symbol), OWL individuals OWL data values; Examples hasparent(?x1,?x2) Λ hasbrother(?x2,?x3) => hasuncle(?x1,?x3)?x = op:numeric-add(3,?z) Student(?x1) => Person(?x1) Note this just duplicates what you could say using OWL subclass statement! Intro to OWL Declan O Sullivan 34 Example of XML Syntax Examples of Atoms Examples <ruleml:imp> <ruleml:_rlab ruleml:href="#example1"/> <ruleml:_body> <swrlx:individualpropertyatom swrlx:property="hasparent"> <ruleml:var>x1</ruleml:var> <ruleml:var>x2</ruleml:var> <swrlx:individualpropertyatom swrlx:property="hasbrother"> <ruleml:var>x2</ruleml:var> <ruleml:var>x3</ruleml:var> </ruleml:_body> <ruleml:_head> <swrlx:individualpropertyatom swrlx:property="hasuncle"> <ruleml:var>x1</ruleml:var> <ruleml:var>x3</ruleml:var> </ruleml:_head> </ruleml:imp> Intro to OWL Declan O Sullivan 35 <swrlx:classatom> <owlx:class owlx:name="person" /> <ruleml:var>x1</ruleml:var> </swrlx:classatom> <swrlx:classatom> <owlx:intersectionof> <owlx:class owlx:name="person" /> <owlx:objectrestriction owlx:property="hasparent"> <owlx:somevaluesfrom owlx:class="physician" /> </owlx:objectrestriction> </owlx:intersectionof> <ruleml:var>x2</ruleml:var> </swrlx:classatom> is a class atom using a class name (#Person) is a class atom using a complex description representing Persons having at least one parent who is a Physician Intro to OWL Declan O Sullivan 36 6

7 Built-Ins Example of Rule using Built-in Built-Ins for Comparisons; e.g., swrlb:equal; swrlb:lessthanorequal; Math Built-Ins e.g., swrlb:add; swrlb:multiply; Built-Ins for Boolean Values e.g., swrlb:booleannot Built-Ins for Strings e.g., swrlb:stringconcat; swrlb:startswith; Built-Ins for Date, Time and Duration e.g., swrlb:date; swrlb:daytimeduration; Built-Ins for URIs e.g., swrlb:resolveuri; swrlb:anyuri Built-Ins for Lists e.g., swrlb:listintersection ; swrlb:member; <ruleml:imp> <ruleml:_rlab ruleml:href="#golddiscount"/> <owlx:annotation> <owlx:documentation>gold customers get a 10% discount on purchases of $500 or more</owlx:documentation> </owlx:annotation> <ruleml:_body> <swrlx:individualpropertyatom swrlx:property="&ex;#hasstatus"> <ruleml:var>customer</ruleml:var> <owlx:individual owlx:name="&ex;#gold"/> <swrlx:datavaluedpropertyatom swrlx:property="&ex;#hastotalpurchase"> <ruleml:var>customer</ruleml:var> <ruleml:var>total</ruleml:var> </swrlx:datavaluedpropertyatom> <swrlx:builtinatom swrlx:builtin="&swrlb;#greaterthanorequal"> <ruleml:var>total</ruleml:var> <owlx:datavalue owlx:datatype="&xsd;#int">500</owlx:datavalue> </swrlx:builtinatom> </ruleml:_body> <ruleml:_head> <swrlx:datavaluedpropertyatom swrlx:property="&ex;#hasdiscount"> <ruleml:var>customer</ruleml:var> <owlx:datavalue owlx:datatype="&xsd;#int">10</owlx:datavalue> </swrlx:datavaluedpropertyatom> </ruleml:_head> </ruleml:imp> use of built-ins to perform comparisons necessary to compute a discount Intro to OWL Declan O Sullivan 37 Intro to OWL Declan O Sullivan 38 Summary An OWL document contains the following kind of information: Class hierarchy info: an OWL document defines class/subclass relationships. Synonym info: an OWL document identifies equivalent classes and equivalent properties. Class association info: an OWL document maps one or more classes to one or more classes, via a property (i.e., domain/range info). Property metadata info: an OWL document contains a lot of metadata for properties. Class definition info: an OWL document specifies the composition of classes. Type of default inferencing/reasoning possible dependent on species of OWL used Additional sophisticated inferencing/reasoning possible through traditional AI languages or through emerging SWRL References [DARPA 2003] Tutorial by Costello and Jacobs of MITRE funded by DARPA OWL web site with lots of information In particular the OWL Guide provides a description of OWL, with many examples: Intro to OWL [DARPA Declan 2003] O Sullivan 39 Intro to OWL Declan O Sullivan 40 Symmetric: if P(x,y) then P(y,x) inverseof: if P1(x,y) then P2(y,x) Transitive: if P(x,y) and P(y,z) then P(x,z) Functional: if P(x,y) and P(x,z) then y=z InverseFunctional: if P(x,y) and P(z,y) then x=z allvaluesfrom: P(x,y) has y=allvaluesfrom(c) somevaluesfrom: P(x,y) has y=somevaluesfrom(c) hasvalue: P(x,y) and y=hasvalue(i) cardinality: cardinality(p) = n mincardinality: mincardinality(p) = n maxcardinality: maxcardinality(p) = n equivalentproperty: P1 = P2 intersectionof: C = intersectionof(c1, C2, ) unionof: C = unionof(c1, C2, ) complementof: C = complementof(c1) Summary of OWL Capabilities (by Roger Costello and Rob Simmons) InverseFunctional P x y 1 oneof: C = oneof(i1, I2, ) Properties: P, P1, P2, P3, equivalentclass: C1 = C2 Generic classes: C, C1, C2, C3, disjointwith: C1!= C2 Specific classes: x, y, z sameindividualas: I1 = I2 Instances: I, I1, I2, I3, differentfrom: I1!= I2 A non-negative integer: n AllDifferent: I1!= I2, I1!= I3, I2!= I3, P(x,y) is read as: property P relates Thing: Intro C1, to OWL C2,, I1, I2,, P1, P2, Declan O Sullivan domain x to range y 41 x Functional P InverseFunctional & Functional P x y 1 1 LEGEND 1 y 7