Easing Participation in the Semantic Web

Easing Participation in the Semantic Web

Stefan Haustein, J ¨org Pleumann

Computer Science VIII, X University of Dortmund,

Baroper Str. 301 D-44221 Dortmund, Germany

f

stefan.haustein, joerg.pleumann

@udo.edu

ABSTRACT

Although a promising idea, the Semantic Web currently

seems to have a problem duplicating the success story of

itspredecessor, theWorldWideWeb. Thenumberofpeo-

pleactivelyparticipatingintheSemanticWebhasbeenvery

limiteduntilnow,becausepeoplecan'tseethebenetsorigi-

natingfromtheextraeorttheyinvestintosemanticallyrich

webpages.Unfortunately,thisadvantageisbarelyvisibleat

alluntilacriticalmassofRDF-annotatedpagesisavailable

onthe net, thus making is diÆcult torecruit new partici-

pantsfortheSemanticWeb. Thearticletriestobreakthis

viciouscircle by showing that the useof appropriatetools

may bothease participation inthe semantic weband pro-

videanumberofadditional advantagesnotdirectlyrelated

to the Semantic Web. The latter, inparticular, may con-

vince a larger number of people to participate, and thus

bringtheSemanticWebneareritscriticalmass.

1. INTRODUCTION

TheSemanticWebisagreatidea. Yet, itdid notquite

takeountilnow. Why isthis thecase? Somearguethat

RDF[19],thelanguageforaddingthesemanticinformation

toexistingwebpagesistheproblem. ThesecriticsseeRDF

as being too complicatedor under-specied[11, 6]. While

RDFtrulyhasits problemsinsomeareas, we don'tthink

that the language itself is the main obstacle that hinders

peoplefromparticipatingintheSemanticWeb. Buttond

outwheretheproblemactuallylies,werstneedtotakea

stepback and look at what made the original websucha

tremendoussuccess.

Inouropinion,therewerefourimportantreasonsforthe

successoftheWorldWideWeb:

Simplicity. HTMLwaseasily understoodandquickly

writtendown. Evennovicescould designafewbasic

web pageswith little eort, putthem ina matching

directorystructureandstartanHTTPdaemontode-

liverthecontenttoclients.

Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission by the authors.

Semantic Web Workshop 2002 Hawaii, USA Copyright by the authors.

Immediatefeedback. AfteranHTMLpagehad been

designedinatext-editor,theresultcouldbedisplayed

inanyHTMLclienttogetanimpressionoftheresults.

Thus, the userhad an immediatefeedback onhis or

herwork.

Additional benets. Even though their original pur-

pose was to present information to other people,

HTMLpagescouldbeusedasameansofdiscussionor

documentationforpeopleparticipatinginaprojector

evenfor personal use. Thus, therewasanadditional

gain users got from participating in the world wide

web, whichmadethe system evenmoreattractive to

them.

Low critical mass. As anetworkedeort, the World

Wide Web required a minimum (but large enough)

number of participants to raise the interest of out-

sidepeople,convincingthemtobecomeinvolved.Yet,

sincetheWorldWideWebwastherstsystemofits

kind,andtherewasnosimilarsystemtocompetewith,

thiscriticalmasswasrelativelylow.

WhenwecomparethesepointstotheSemanticWebinits

currentform,wenoticethatmostofthemarenotfullled:

Simplicityisonlypartiallygiven.ThemixtureofRDF

and DAML+OILis understoodinall its detailsonly

by people that have a background in AI or related

elds. Noviceswillonlybeabletouse basicconcepts

ofRDFandmightthushaveproblemstoseethereal

advantagesoftheSemanticWeb.

Immediatefeedbackisnotgiven.Unfortunately,there

isnospecicclientsoftwarefortheSemanticWebthat

givesusersanimpressionoftheirRDFfactbase. One

couldarguethatitdoesn'tevenmakesensetoaskfor

sucha software, because the clients of the Semantic

Webareprogramsratherthanhumanbeings.

There are no additional benets, at least none that

are ovious to "`ordinary end-users"'. While human-

readable HTML pages primarily designed for other

people can also be used for personal purposes, this

isnottrueforRDFfacts,whicharemeanttoberead

byprograms.

Thecritical massis considerablyhigher. Whyisthis

the case? This time, there already is an existing

system | the original World Wide Web | , and

most peoplenowadaystendto usethe"`bruteforce"'

method to nd a specic piece of information in it,

namelyGoogleorsomeothersearchengine. Thus,it

ismorediÆculttoconvincepeopletotakepartinan-

othersystem,evenifitisanextensiontotheexisting

one.

Aslongastherstthreepointsaretrue,thecriticalmass

ofusersneededtomaketheSemanticWeb"`takeo"'will

behardtoreach.Unfortunately,seen theotherwayround,

theSemanticWebhardlyhassomekindofrealbenetunless

thereis alarge-enoughnumberofparticipants that makes

availableRDF-speciedinformationtoothers,thatis,until

thecriticalmassisreached. Thecurrentsituationcouldbe

seen as some kind of vicious circle that has to be broken

beforetheSemanticWebhasachancetosucceed.

2. TOOLS TO BREAK THE CIRCLE

Tobreakthecircle,wehavetogetridofasmanyaspos-

sible of the four problems shown in the previous section.

Sincewecannotlowerthecriticalmass formainstreamac-

ceptance of the Semantic Web (possibly by forcing people

intoit),wehavetofocusontheotherthree: Simplicity,im-

mediatefeedback,andadditionalbenets. Averypromising

waytoachievethisseemstobetheuseofappropriatetools.

ThesetoolswouldhavetoeaseparticipationintheSemantic

Web,butwouldalsohavetoprovide some"`addedvalue"'

thatmakesthemattractiveto end-users. Obviously,when

usingthetools,peoplewillalsolikelyparticipateinthese-

mantic web, even if that is not their original motivation.

Thefollowingsectionstrytoshowwhatfeaturesthesetools

mightoer.

2.1 Generative approach

Lookingat existingtools developedfor or related tothe

SemanticWeb,forexampleProtege-2000[23]orOntobroker

ontology and fact management. Information is storedina

knowledgebaseprovidingnegrainedaccess. Theontology

isutilizedtomakesurethatthecontentcorrespondstothe

desired structure. The two systems mentionedabove are

abletoexporttheirfact basetoanRDFrepresentation.

Whilethesetools aim into theright direction,theystill

have aproblem: Aslong as onewants amachine-readable

RDF-version of the facts as well as a human-readable

HTML-version,duplicateeortisrequiredtomaintainboth.

Take,forexample,atypicalwebsiteforauniversitydepart-

ment containing information aboutthe department'ssta,

researchtopics,projects,and publications. Ahighlystruc-

turedsitelikethisissuitableforparticipatingintheseman-

ticweb,anditcaneasilybemodelledusingacorresponding

domain ontology. Yet, a change as simple as a telephone

numberhastobepropagatedtotheRDFversionaswellas

theHTMLversion.

Given a Semantic Web tool followed a generative ap-

proach,thesituationwouldbeeasier: Assumethistoolwere

abletoincorporateregularHTMLfortheunstructuredpart

ofthewebsite,andthesepagescouldcontainplaceholders

forinsertionofinformationcontainedinthefactbase. The

toolwouldthenbeabletogeneratetheactualHTMLpages

automaticallyfromtheexistingRDFinformation{oreven

both from acommonfact base { ,thus requiringthe user

tomaintainthisfactbaseonly,atleastasfarasstructured

information is concerned. If the generation ofpages takes

placeatrun-time,wearriveat atoolthatcouldbeseen as

a"`SemanticWeb-enabledHTTPserver"'

Whiletheavoidanceofredundancyalreadyisabigadvan-

tageaddressingsimplicity,thegenerativeapproachprovides

otheradvantagesthatfallintotheareaof"`addedvalue"':

Incontrast to editing HTMLdirectly, aunique look

and feel can easily be established for the whole site,

givenanappropriatetemplatemechanism.

factbase,loweringredundancyevenfurther.

Incontrast toplainHTMLles, ontology-basedcon-

sistency checkscanbeperformedautomatically while

entering data,e.g. avoidingdangling linksinsidethe

system.

2.2 Incorporation of database features

To broaden the possible target audience of our Seman-

tic Web server, we might try to incorporate database-like

featuresandthuspositionitasanalternativetoa"`heavy-

weight"'databasesolution.

Whilerelationaldatabases withHTML-generating front-

endarequitecommonthesedays(e.g. ColdFusion[8],PHP

[2], Enhydra [1] etc.), these solutions are mainly used for

siteswithasimple,low-dimensionalstructure,suchasguest

books or news pages (e.g. Slashdot.org). More complex

domainssuchasuniversitydepartmentsoftenstilluseplain

HTMLlesfortheirwebpresentation,ormakeonlylimited

useofdatabasetables.

Here, the reason may be that a high number of ta-

bleswouldberequiredformodelingevensimpleontologies,

mainlybecause associations are not rstclass membersof

relational database systems. Revisiting the university de-

partment scenario, weneed atleast tables for persons,re-

search topics, projects, and publications. Figure 1 shows

apossible UMLclassdiagramofthedatabase'sconceptual

model. Since all n:n associations require separate associa-

tiontables, this results in quite alot of normalised tables

(more than10), each of which potentiallycontains onlya

verysmallsubsetofall thepossibleinstances.

Inthis case, thebenetfor the creator, that is, the dy-

namicgenerationofHTMLor{inourcase {RDFfroma

singlesetofdata,doesnotoutweightheextraeortinherent

inmaintainingthetables.

UsingSemanticWebtools,thepicturemaychangesignif-

icantly. Foralownumberofinstances, theinternalknowl-

edgebase providedby aSemantic Web tool may be suÆ-

cient. Associationsare direclysupported,andtheontology

languagealsoallowstospecifyintegrityconstraintsforthem

at anappropriate level. SinceSemanticWebtools usually

comewithagenericuserinterface,theneedtocreateHTML

formsforeditingthetablesisavoided.

2.3 Incorporation of Content Management Features

AnotherareathataSemanticWebtoolmight addressis

content management. Content managementsystems, such

as Hyperwave [3], Zope [5] or OpenCMS [4] provide user,

versionandmetadatamanagementforasetofHTMLpages

orbinarydocumentsinotherformatssuchasPDForWord.

Theirset ofmetadata, hoewever, is usuallyxedand tai-

loredto themostcommonneeds. Here,ontology-basedSe-

manticWebtoolsprovidemuchmoreexibility,andmaybe

superiortogeneralcontentmanagementsystemsindomains

where themeta datarequirementssignicantlydierfrom

thestandardsetprovidedbycontentmanagementsystems.

2.4 Openess to Alternative Schema Languages

Intheintroduction,weclaimedthatbeneathprovidingno

gainthatbecomesimmediatelyobvious,RDFannotationis

learnyetanotherformaldescriptionlanguage. Usershaving

anbackgroundinAImaybeexpectedtobefamiliarwithde-

scriptionlogicsandcorrespondingontologymodellingtools.

For mainstream acceptance, though, integration of recog-

nisedstandardslikeUML[20]mayhelp toimproveaccep-

tance of Semantic Web tools and thus lowerthe entrance

barrier [13]. Most studentsof computerscience orrelated

engineering disciplines canbe assumedto befamiliar with

UML and modellingtools like Together or Rational Rose.

These students could easily apply their modelling knowl-

edgetotheSemanticWebandthuscontributetoitsgroup

ofearlyadopters.

3. THE INFORMATION LAYER

InordertodemonstratethatparticipationintheSeman-

ticwebactuallycanbesimple,andthatusingaserverbased

onanegrainedfactbaseinstead ofHTML-or RDFles

canprovideimmediategains, wehavestartedtomodelour

ownunit'swebpagesaccordingly. Forthispurpose,weused

ourInformationLayersystem,whichstoresdatainasimple

XML formatthat is determinedby agivenontology. The

informationlayerusesanobject-orientedmodelfordatarep-

resentation. Objects consistof atomicattributesand rela-

tionstootherobjects. Theconsistencyofrelations inboth

directionsisensuredautomatically,avoidinginconsistencies

inside thesystem. Theconcepts and relations are dened

application-dependentinanexternalontologydenitionle.

All les used by the information layerare stored as XML

documents.

TheInfoLayersystemwasoriginally designedasaninte-

gratedinformationplatformforsoftwareagentsandhuman

usersinaconference scenario. Thesystemwasusedinthe

COMRISproject[21]inordertomakeconferenceinforma-

tionavailableinappropriateformatstohumanusersaswell

assoftwareagents,utilizingthesameunderlyingknowledge

base. AccesstothecontentispossibleviaagenericHTML

interface as wellas a FIPA [16] basedXML interface [18].

Obviously,when information is machine readable for soft-

ware agents, it is not a big leap to make this information

availablefortheSemanticWebaswell.

Intheprocessofmodellingourunitwebpages,wemade

severalimprovementstooursystem,simplifyingtheuse as

areplacementfor a\regular"webserver. Whiletheremay

bealternativepathsappropriateforothersystems,ourmain

purposewastoshowthatusingsemanticwebsystemsmay

provide direct advantages over regular web servers, even

withoutrelyingonadvancedfeaturessuchasknowledgeinte-

grationfromdierentsources(e.g.KAON-REVERSE[17]).

3.1 XMI Import

TheoriginalversionoftheInformationLayersystemused

its own proprietary XML-based ontology description lan-

guage. Inordertosimplifytheinitialstepofgeneratingthe

application ontology,we havereplaced the internalformat

byXMI[20],theXMLbasedexchangeformatforUMLdi-

agrams. Figure 1 shows a simplied version of the UML

modelcurrentlyusedasabasisfor ourunitwebpages.

WehavechosenUMLasontologymodellinglanguage[13]

instead ofRDFS[7] because itis diÆcultto avoid contact

withUMLwhenworkingincomputerscienceorintheITin-

dustryingeneral. Formostcomputerscientists,aUMLed-

tool box. Thus, the extra eort of installing and getting

familiar with an RDFS editor, possibly preventing people

fromgettingintouchwiththeSemanticWeb,isavoided.

Comparedto otherlanguages suitablefor ontologymod-

elling, UMLcurrentlystill lacksclearly dened semantics.

However,therearesignicanteortstosolvethisproblems

[22,10].

Thisaspectmaybelessimportant forsystemsproviding

theirowncomfortableOntologyeditor.

3.2 HTML Generation

Themostimportantcapabilityrequiredforbeingableto

replaceexisting webserversis {ofcourse{thegeneration

ofHTMLpages.

The information layer contains a module that provides

built-inweb-serverfunctionality. Theserverisable togen-

erate HTML dynamically: For any object, the attributes

aresimplydisplayed,and theassociationsto otherobjects

are convertedto sets of hyperlinks to the related objects.

Conceptsaredisplayedasaclickablelistofinstancescorre-

spondingtothe concept. TheHTMLinterfacecanalsobe

usedtoeditthecontentofthesystemusingformsgenerated

dynamicallybasedontheontology. IntheCOMRISproject,

theHTMLinterface was usedfor interactionwiththe end

useraswellforasdebuggingandinspectionpurposes.

InadditiontogenericHTMLgeneration,templatescanbe

usedinordertogenerateHTMLpagesconformingtoagiven

look andfeel. Inthe COMRISproject, we have alsoused

thetemplatemechanismtogeneratetheinputstructurere-

quiredbythetextgenerationsystemTG/2([9])whichwas

usedtogeneratenaturallanguageoutputforawearablede-

vice. Thetemplate mechanism is described in somemore

detailinthenextsection.

3.3 SWRC and RDF Integration

The Semantic Web Research Community (SWRC) On-

tology [24] is an ontology designed in order to describe

the structure of the Semantic Web Research Community,

namely the members, events, topics and projects, in a

machine-readable manner. It is available in DAML+OIL

tancehierarchyoftheSWRContology.

Sinceour\local"researchunitontologywasprimarilyde-

signedtottheneedsofour\regular"webpresentation,it

doesnotmatchthe\shared"SWRContologyexactly.How-

ever, usingthe template mechanism ofoursystem, weare

able to generate RDF pages corresponding to the SWRC

ontology on the y. Figure 3 shows a simplied example

template that is used to generate SWRC-compliant RDF

content for instances of the class \Member". Inthe tem-

plates, elements ina special namespace, denotedby the t

prexintheexample,arereplacedbycontentqueriedfrom

the InformationLayerwithrespectto thecurrent instance

whichisdeterminedfromthepageURL.

Thus, it is possible to participate in the Semantic Web

without needing to extend a predened shared ontology,

whichmaybebloated andstillnotfullll alllocalrequire-

ments. Instead,thedomainofinterestcanbemodelledus-

ingaleandomainspeciclocalontology. TheSWRCperson

nameslotillustratestheadvantageofthisapproach:SWRC

containsonlyonepersonnameslotthatisnotsplitintorst

andlastname. Ifthelocalapplicationrequireshavingboth

partsavailableseparately,itwouldbenecessarytoduplicate

thecorrespondinginformation,whenbuildingthelocalon-

tologyontopoftheSWRContology.Also,SWRCconcepts

like\Organization"maynotberequiredinalocalontology

coveringasingleorganization. Informationaboutthelocal

organization can bestored inasingle staticRDF le, not

bloatingthelocalontology.

InadditiontotemplatebasedRDF generation,itwould

be possible togenerateRDFdirectlycorrespondingto the

local ontology automatically [12]. However, this feature is

notimplementedyet.

3.4 Infrastructure Integration

Forsimpler integration withthe existing Web serverin-

frastructure, we changedtheInformation Layerimplemen-

tation to become a Java Servlet instead of a stand alone

program. RunningtheInformationLayerasaJavaServlet

allowssmoothintegrationwithexisting Webpresentations,

withoutanyhardswitch. Theservice cansimplybeadded

Servlet-Container (e.g. Tomcat) Servlet-Container (e.g. Tomcat) Ontology

(UML-XMI) Fact base (XML)

Infolayer

Template based XML generation

<rdf:RDF xmlns:t="http://infol

<swrc:ScientificStaff rdf:ID

<swrc:name>

<t:eval expr="givenName+

</swrc:name>

<swrc:email>

<t:eval expr="email" />

</swrc:email>

Templates

<RDF xmlns="http://www.w3.org/1999/02/22-

<ScientificStaff

xmlns="http://www.semanticweb.org/ont xmlns:p0="http://www.w3.org/1999/02/2

<name>Stefan Haustein</name>

<email>[email protected]</email>

<phone>+49 231 755 2499</phone>

<fax>+49 231 755 5105</fax>

</ScientificStaff>

</RDF>

XHTML RDF

Figure3: DynamicHTMLandRDFgenerationusingtheInformation Layertemplatemechanism. Elements

prexedwithat: areevaluatedwithrespecttothe currentinstanceasgivenin thepage URL.

3.5 File Upload

Lastbutnotleast,wehavetheoptiontouploadarbitrary

Files(PDF,MPG,...) intothesystem. Wehaveaddedthis

feature in order to improve suitability for general useage.

Whileitmaylook abitoddhere onthe rstsight, it isa

typicalfeatureofcontentmanagementsystems. Ofcourse,

thecontentofthelesisopaquetothesystem,whichiscon-

troversial to theidea ofprovidingnegrained information

inRDF-format.However,thesystemsupportstheaddition

ofrelevantmeta-information.

3.6 Installation

Acomplicatedinstallation proceduremaypreventpoten-

tial users from actually using a system, even if there are

obvious timeor cost savings inthe long run. Building on

thesystem improvementsdescribedabove,the installation

of anInformation Layer based system was reduced to the

followingsteps:

1. Build a simple base ontology with the UML tool of

yourchoice,or justusethesampleontologyavailable

fromtheinfolayerwebpageasastartingpoint.

2. InstallApache-TomcatoranyotherWebserverthatis

capableofhandlingJavaservlets,ifnotalreadyavail-

able.

3. Install theInformationLayerServletlesinthe Web

servicesdirectoryoftheserverandadoptthecongu-

rationintheweb.xmlletoyourlocalenvironment.

Followingthesesteps,auserisalreadyabletoaddcontent

usingthegenericWebinterfaceandtoviewthecontentus-

ingthatinterface. Nowthesystemcanbefurtherenhanced,

byextendingtheontologyandbyaddingXHTMLandRDF

templates,customisingthelookandfeelandtheRDFgen-

erationpropertiesofthesystem.

Pleasenotethatthetemporalframeofthelattertwosteps

is not xed. For example, onecould start with managing

publicationsusingthesystem,andthenlateraddothercon-

ceptslikeprojects,topics,personsorcourses.

3.7 Related Systems

Obviously, other Semantic Web tools may be extended

similarily. Protege-2000 is a Knowledge Base supporting

RDFformat. ItprovidesaniceJavauserinterfaceincluding

anontologyeditor,butcurrentlylacksaplainHTMLinter-

face. Other semantic web tools suchas Ontology Builder

[14] or the KAON framework [17] seem to focus more on

enterprise-levelontologymanagementandinformationinte-

gration.

4. CONCLUSION AND OUTLOOK

TheSemanticWebisagreatvision. However,forabroad

adoption, simple tools that allow participation without a

backgroundin AI are still rare. Protege and similartools

seemto aiminthisdirection. Wewouldliketo contribute

ourowntool,theInformationLayer. Whileothertoolsfocus

oneasingtheontologybuildingprocess,wemainlytriedto

addresssimplicityintheoverallsystemaswellasproviding

tionofHTMLaswellasRDFfromacommonfactbaseto

avoidredundancy,othersaretheincorporationofdatabase

andcontentmanagementfeaturestobroadenthetargetau-

dience. Thisway,wehopetoimproveavailabilityofstruc-

tured information suitable for the Semantic Web. We did

notputafocus onadvancedfeatures likefull DAML+OIL

support,nordonothaveapriorityhereinthefuture.

AwebsitethatutilizestheInformation Layerinitscur-

rent formis adatabasefor Java-enabledsmall deviceslike

cellphonesandpersonaldigitalassistants 1

. Here,theontol-

ogy descibes the devices, their capabilities, vendors, avail-

able protocols and knownbugs. Changesto the fact base

are quitefrequent, butdonotrequiretheduplicatedeort

ofupdatingahumanandamachine-readableversion,which

makesthesiteveryeasytomaintain.

TheInformationLayerisalsobeingusedasaprototypical

webpresenceforMuSofT,aGermany-wideprojectthatde-

velopsmultimediateachingmaterialforsoftwareengineering

education. Thesite'sgoal is tomanageand distribute the

learningobjectscontributedbythevariousprojectpartners.

Thisinstallationmakesuseofthecontentmanagementfea-

turesthesystemprovides: Learningobjectcanbeuploaded

into thesystemfromaWebbrowser. ToalloweÆcient re-

trievalofmaterial,LOM 2

-conformingmeta-dataisprovided

usingthesystem'sontologycapabilities.

Previous versions ofthe Information Layersystem have

beenandstillareareusedasabasisfortheMLnetteaching

informationserver 3

andinvariousinternalprojects.

FormoredetailsabouttheInformationLayeranditsap-

plications,pleaserefertohttp://infolayer.org.

5. REFERENCES

[1] Thehomeofenhydra.org,2001.

http://www.enhydra.org.

[2] Php:Hypertextpreprocessor-homepage,2001.

http://www.php.net.

[3] Hyperwavehomepage,2002.

http://www.hyperwave.com.

[4] Opencmshomepage,2002.http://www.opencms.org.

[5] Zopehomepage,2002.http://www.zope.org.

[6] DanBrickley.RDFinterestgroup-issue tracking.

Technicalreport,WorldWideWebConsortium,2000.

http://www.w3.org/2000/03 /rdf-tracking/ .

[7] DanBrickleyandR.V.Guha.RessourceDescription

Framework(RDF)Schemaspecication1.0. Technical

report,WorldWideWebConsortium,2000.

http://www.w3.org/TR/CR-rdf-schema-20000327.

[8] RobBrooks-Bilson. ProgrammingColdFusion.

O'Reilly,2001.

[9] StephanBusemann.Ashallowformalismfordening

personalizedtext.InWorkshopProfessionelle

Erstellung vonPapier-und Online-Dokumentenatthe

22ndAnnual GermanConferenceonArticial

Intelligence(KI-98),Bremen,September1998.

[10] T.Clark,A.Evans,S.Kent,S.Brodsky,andS.Cook.

AfeasibilitystudyinrearchitectingUMLasafamily

oflanguagesusingapreciseOOmeta-modeling

1

http://www.kobjects.org/devicedb

2

http://ltsc.ieee.org/wg12/index.html

3

2000.http://www.cs.york.ac.uk/puml/mml/mmf.pdf.

[11] WolframConen,ReinholdKlapsing,andEckhart

Koeppen.RDFmodelandsyntaxrevisited: From

reicationtonesting,fromcontainerstolists,from

dialecttopurexml.InInternationalSemanticWeb

WorkingSymposium(SWWS),pages195{208,

StanfordUniversity,California,USA,2001.

[12] S.Craneeld.Networkedknowledgerepresentation

andexchangeusingUMLandRDF.JournalofDigital

information,1(8),2001.

[13] S.CraneeldandM.Purvis.Umlasanontology

modellinglanguage.InProceedings oftheWorkshop

onIntelligentInformationIntegration, 16th

InternationalJointConferenceonArticial

Intelligence(IJCAI-99,1999.

[14] AseemDas,WeiWu,andDeborahL.McGuinness.

Industrialstrengthontologymanagement.In

InternationalSemanticWeb WorkingSymposium

(SWWS),pages17{37,StanfordUniversity,

California,USA,2001.

[15] StefanDecker,MichaelErdmann,DieterFensel,and

RudiStuder.Ontobroker: Ontologybasedaccessto

distributedandsemi-structuredinformation.In

R.Meersmanandother,editors,Semantic Issuesin

MultimediaSystems, KluwerAcademic Publisher,

Boston,1999.KluwerAcademicPublisher,Boston,

1999.

[16] FoundationForIntelligent PhysicalAgents(FIPA).

FIPAAgentManagementSpecication,2000.

http://www.pa.org/specs/pa00023/XC0 002 3F. pdf.

[17] SiegfriedHandschuh,AlexanderMaedche,Ljiljana

Stojanovic,andRaphaelVolz.KAON-theKArlsruhe

ONtologyandSemanticWebinfrastructure.

http://kaon.semanticweb.org,2002.

[18] StefanHaustein.Utilisinganontologybased

repositorytoconnectwebminersandapplication

agents.InProceedingsof theECML/PKDDWorkshop

onSemanticWeb Mining,September2001.

[19] OraLassilaandRalphR.Swick,editors.Resource

DescriptionFramework (RDF)ModelandSyntax

Specication.WorldWideWebConsortium,1999.

http://www.w3.org/TR/1999/ REC-rdf-syntax-

19990222.

[20] ObjectManagementGroup.OMGUniedModeling

LanguageSpecication,version1.3.

http://www.omg.org/technology/documents/formal/

uniedmodelinglanguage.htm,2000.

[21] EnricPlaza,JosepLlusArcos,Pablo Noriega,and

CarlesSierra.Competingagentsinagent-mediated

institutions.PersonalTechnologies Journal,2(3):1{9,

1998.

[22] PreciseUMLGroup.ThePreciseUMLGrouphome

page.http://www.puml.org,2001.

[23] StanfordUniversity.UsingProtege-2000 toEditRDF,

June2000.

http://protege.stanford.edu/protege-rdf/protege-

rdf.html.

[24] YorkSure.Swrc-semanticwebresearchcommunity

ontology,2001.

http://ontobroker.semanticweb.org/ontos/swrc.html.