129
JeenBroekstra,Michel Kleinetal.:Addingformal semanticstothe Web
9
Table2.Slot-definitionsinOILandthecorrespondingRDF(S)constructs.
OILprimitive
RDFSsyntax
type
slot-def
rdf:Property
class
subslot-of
rdfs:subPropertyOf
property
domain
rdfs:domain
property
range
rdfs:range
property
inverse
oil:inverseRelationOf
property
transitive
oil:TransitiveRelation
class
symmetric
oil:SymmetricRelation
class
5 Conclusion
In the previous section we have shown that it is possible to
define a formal knowledge representation schema as an ex-
tension to RDFS, effectively implementing the ”third layer
ofthe Semantic Web”. We didthis by defining the ontology
language OIL in RDFS, using existing primitives as much
as possible whileretainingaunambiguousmappingbetween
theoriginalOILspecificationanditsRDFSserialization.The
resulting extension of RDFS allows the specification of do-
main ontologies that are alreadypartially understandable by
non-OIL-aware RDFSapplications, while OIL-aware appli-
cationscanfullybenefitoftheaddedfeatures,suchasformal
semantics andreasoningsupport.
There are still afew unsolved problems withthespecifi-
cationofOIL intoRDFS. First, wedidnottake intoaccount
arestrictionontherdfs:subClassOfstatement,i.e. therestric-
tion that no cycles are allowed in the subsumption hierar-
chy. Wethinkthatthisrestrictionshouldbedropped:without
cycles one cannot even represent equivalence between two
classes — in our view this is an essential modeling primi-
tive for any knowledge representation language. Moreover,
these kinds of constraint significantly add to the complex-
ityof parsing/validatingRDFdocuments in away whichwe
think would be highly undesirable. This is because they are
really semantic constraints rather than syntactic ones (they
limitthekinds ofmodels thatcanberepresented),evenifthe
reasoningrequiredinordertodetectconstraintviolationisof
averybasic kind.
Second, in contrast with RDFS, OIL allows more than
one range restrictionona property. Althoughthis canbecir-
cumventedbydefining a dummysuperclass ofallclasses in
the range restriction, we see noreason for this restriction in
RDFS. From a modeling point of view, allowing more than
one rangerestrictionis amuchcleanersolution.
During the process of extending RDFS, we encountered
acouple of peculiarities in the RDFS definition itself. The
moststrikingoftheseisthenon-standardobject-metamodel,
asalreadydiscussedinsection2.2.1. Themainproblem with
this non-standard model is that some properties have a dual
role in the RDFSspecification, both attheschema leveland
instance level (cf. [Nejdletal.,2000]). This makes itquite a
challengeformodelerstounderstandtheRDFSspecification.
We tried to make this distinction clear in our extensions by
usingtherdf:typerelationshipconsistentlyas anobject-meta
relationship.
Furthermore, the semantics of several relationships are
unclear. Itis notobviousthatthemeaningofa listofdomain
(orrange)restrictions is the unionofthe elements. Also, the
meaningofthesubPropertyOfrelationwithrespecttothe in-
heritanceofthedomainandrangerestrictions is unclear.
Despite these problems, we think that this procedure of
extendingRDFSisalsoapplicabletootherknowledge repre-
sentationformalisms.
Acknowledgements. We wouldliketothankMonicaCrubezy,Ying
Ding, Michael Erdmann, Frank van Harmelen, Arjohn Kampman,
and BorysOmelayenkofortheir helpful commentsandforreview-
ingearlydraftsofthispaper.
References
Bechhoferetal.,1999. Bechhofer, S., Horrocks, I., Patel-
Schneider, P. F., and Tessaris, S. (1999). A proposal for a
descriptionlogic interface. InProc.of DL’99,pages33–36.
Berners-Lee,1998. Berners-Lee, T. (1998). Semantic web road
map.
Internal note, World Wide Web Consortium.
See
http://www.w3.org/DesignIssues/Semantic.html.
BrickleyandGuha,2000. Brickley, D. and Guha, R. (2000). Re-
source Description Framework (RDF) Schema Specification1.0.
Candidate recommendation, World Wide Web Consortium. See
http://www.w3.org/TR/2000/CR-rdf-schema-20000327.
Fenseletal.,2000. Fensel, D., Horrocks, I., van Harmelen, F.,
Decker,S., andKlein,M.(2000). OILina nutshell. InProceed-
ingsof12thInternationalConference onKnowledgeEngineering
andKnowledgeManagement,Juan-les-Pins,FrenchRiviera.
Horrocksetal.,2000. Horrocks, I., Fensel, D., Broekstra, J.,
Decker,S.,Erdmann,M.,Goble,C.,vanHarmelen,F.,Klein,M.,
Staab, S., and Studer, R. (2000). OIL: The Ontology Inference
Layer. Technicalreport,UniversityofManchester/VrijeUniver-
siteitAmsterdam. Seehttp://www.ontoknowledge.org/oil/.
Kleinetal.,2000. Klein, M., Fensel, D., van Harmelen, F., and
Horrocks,I.(2000). Therelationbetweenontologiesandschema-
languages: Translating OIL-specifications in XML-Schema. In
Proceedings of the Workshop on Applications of Ontologiesand
Problem-solving Methods, 14th European Conference on Artifi-
cial IntelligenceECAI2000,Berlin,Germany.
LassilaandSwick,1999. Lassila, O.andSwick,R.R.(1999). Re-
source Description Framework (RDF): Model and Syntax Spec-
ification. Recommendation, World Wide Web Consortium. See
http://www.w3.org/TR/REC-rdf-syntax/.
Nejdl etal.,2000. Nejdl, W., Wolpers, M., and Capella, C.
(2000). The RDF Schema Revisited. In Modelle und Model-
lierungssprachen in Informatik und Wirtschaftsinformatik, Mod-
ellierung2000,St.Goar.FoelbachVerlag,Koblenz.
Staabetal.,2000. Staab, S., Erdmann, M., M¨adche, A., and
Decker,S.(2000). An extensible approachfor modeling ontolo-
gies in RDF(S). In First Workshop on the Semantic Web at the
FourthEuropeanConferenceonDigitalLibraries,Lisbon,Portu-
gal.
StaabandM¨adche,2000. Staab, S. and M¨adche, A. (2000). Ax-
iomsareobjects,too-ontologyengineeringbeyondthemodeling
of concepts and relations. Technical Report 399, Institut AIFB,
Universit¨atKarlsruhe.