F. An ontology-based mediation architecture for e-commerce applications

O.Corcho1,A.Gomez-Perez1,A.Leger2,C.Rey3,andF.Toumani3

123

UniversidadPolit´ecnicadeMadrid(ocorcho@fi.upm.es,asun@fi.upm.es)FranceTelecomR&D(alain.leger@rd.francetelecom.com)LIMOS,Universit´eBlaisePascal,France(rey@isima.fr,ftoumani@isima.fr)

Abstract.AspartoftheMKBEEMproject,wepresentanontologybasedmedi-ationframeworkforelectroniccommerceapplications.Theframeworkisbasedonamediator/wrapperapproachthatsupportsanintegratedviewovermultiplehet-erogeneoussources.TheMKBEEMmediationsystemallowstofillthegapbetweencustomersqueries(possiblyexpressedinanaturallanguage)anddiversespecificprovidersoffers.Incontrastwithmanyexistingmediatorbasedsystems,ourap-proachrestsonathree-layerknowledgerepresentationarchitecturewhichincludesanelectronicservicesontologybesidestheusualdomainontologyandsourcesde-scriptionslayers.Atthereasoninglevel,weproposeanewmechanism,namelydynamicdiscoveryofe-services,thatactsincollaborationwiththePicselmediatorsystemtoeffectivelyachievetheMKBEEMmediationtasks.

1Introduction

Therecentprogressandwiderdisseminationofelectroniccommerceviatheworldwidewebhasdramaticallyincreasedthediversityandheterogeneitiesincorrespondingsystems.Toavoidthisprogresstobehamperedbyclosedmarketswithincompatibleapplicationsthatcannotuseeachotherservices,thereisaneedforefficientandflexiblemechanismstohandletheinteroper-abilityproblem.Today,weexpectthatemergingstandardswillsolvemostofthesyntacticinfrastructureproblems.However,dealingwithsourcesfromdifferentdomainsandtheirsemanticsdifferencesrequirestohandlethein-teroperabilityproblematthesemanticlevel.

Thenotionofmediationhasbeenproposedastheprincipalmeanstoresolveproblemsofsemanticinteroperation[13].Mediationarchitecturesaregenerallybasedonthemediator/wrapperparadigmwhereinformationsourcesare”wrapped”tologicalviewssothattheirinterfacestotheoutsideworldareuniform[13,8].Theselogicalviewsare”glued”togetherthroughanintegratedglobalschemausuallycalledthedomainontology.Insucharchi-tectures,amediatoractsasaninterfacebetweenuserqueriesorapplicationprogramsandexistinginformationsources:queriesagainstthedomainon-tologyarereformulatedintermsoflogicalviewsandthensenttotheremotesources.

Inthispaper,wepresentathree-layerontology-basedmediationframeworkforelectroniccommerceapplications.Incontrastwithmany

2O.Corchoetal.

existingmediatorbasedsystems,theproposedarchitectureincludesanelec-tronicservicesontologybesidestheusualdomainontologyandsourcesde-scriptionslayers.Roughlyspeaking,anelectronicservice(e-service)1canbeseenasanapplicationmadeavailableviaInternetandaccessiblebyhumansand/orotherapplications[2,3,6].Examplesofe-servicescurrentlyavailablerangefromweatherforecast,on-linetravelreservationorbankingservicestoentirebusinessfunctionsofanorganization.Whereasinmanye-commerceplatforms,e-servicesareassociatedtoproviders,weproposetodefineane-serviceasanintegratedprovider-independentofferavailableonagivene-commerceplatform.Then,toeffectivelyhandlethemediationtasks,werelyontworeasoningmechanisms:

•thefirst,calleddynamicdiscoveryofe-services,allowstoreformulateusersqueriesagainstthedomainontologyintermsofe-services.Theaimhereistoallowtheusersandapplicationstoautomaticallydiscovertheavailablee-servicesthatbestmeettheirneedsatanygiventime,

•thesecond,calledqueryplangeneration,allowstoreformulateauserquery,expressedasacombinationofe-services,intermsofprovidersviews.Theaimofthissecondissueistoallowtheidentificationoftheviewsthatareabletoanswertothequery(knowingthat,afterwards,thequeryplanswillbetranslatedintodatabasesqueriesviathecorrespond-ingwrappers).

Whilethesecondreasoningmechanism,knownasqueryrewritingusingviews,hasalreadybeenaddressedintheliterature[11,5,8],thefirstisaquitenewproblemforwhichweproposeasolutioninthispaper.

Therestofthispaperisorganizedasfollows.Thecontextofthisstudy,namelytheMKBEEM2project,isintroducedinSection2.ThearchitectureoftheMKBEEMontologyispresentedinSection3.InSection4,thedynamicidentificationofe-servicesisdetailed.WeconcludeinSection5.

2TheMKBEEMmediationsystem

TheglobalaimoftheMKBEEMprojectistoextendcurrentelectroniccom-merceplatformstoreachatrulypanEuropeanandculturallyopenelectroniccommercemarket.ThemaintechnicalaimofMKBEEMistocreateanin-telligentknowledgebasedmultilingualmediationservicewhichdisplaysthefollowingfeatures:

•Naturallanguageinterfacesforboththeenduserandthesystem’scon-tentproviders/serviceproviders.

•Automaticmultilingualcataloguingofproductsbyserviceproviders.

12

Alsocalledwebservices.

MKBEEMstandsforMultilingualKnowledgeBasedEuropeanElectronicMar-ketplace(IST-1999-10589,1stFeb.2000-1stAug.2002).

Anontology-basedmediationarchitecture3

•On-linee-commercecontractualnegotiationmechanismsinthelanguageoftheuser,whichguaranteesafetyandfreedom.AdetaileddescriptionoftheMKBEEMprojectandrelevantreferencesrelatedtothisprojectcanbefoundin[1].

Letusconsideranexampleofasimplifiednaturallanguagerequestsce-nariotrainedonthegenericMKBEEMarchitecturetosketchtherolesofthedifferentcomponentsoftheMKBEEMsystem:anendusersubmitstotheMKBEEMsystemanaturallanguagequery.ThequeryisreceivedbytheUserAgentcomponentwhichrecognizestheuserlanguageandthenforwardsittothecorrespondingHumanLanguageProcessingServer(HLPServer).TheHLPServerisinchargeofmeaningextraction:heanalyzestheinputstringandconvertsthequerytoanOntologicalFormula(OF)whichisalanguage-independentformulacontainingthesemanticinformationofthecorrespond-ingphraseofhumanlanguage.TheOFissenttotheRationalAgent(RA)whichisresponsibleofthedialoguemanagementbetweenthedifferentcom-ponentsoftheMKBEEMsystem.TheRAforwardstheOFtotheDomainOntologyServer(DOS).TheDOSmoduleisresponsibleofstoring,access-ingandmaintainingtheontologiesusedbytheMKBEEMsystem.Italsoprovidesthecorereasoningmechanismsneededtosupportthemediationservices.Continuingwiththeexample,theDOSachievesacontextualinter-pretationoftheformulausingitsknowledgeabouttheapplicationdomain.Thistaskconsistsmainlyintheidentificationoftheoffers(e-services)deliv-eredbytheMKBEEMplatformthat”bestmatch”theontologicalformula.ThesetofsolutionscomputedbytheDOSissentbacktotheRAwhich,incollaborationwiththeUserAgent,willasktheusertochooseonesolutionandtocomplete,ifany,theparametersthataremissing.Afterthisdialoguephase,theretainedsolutionissentbacktotheDOStogeneratethequeryplans.AqueryplancontainsinformationabouttheContentProvidersAgents(CPA)thatareabletoanswertotheuserquery.Then,thankstospecificwrappersbelongingtoCPAaqueryplanistranslatedintodatabasesquerieswhichareexecutedontheremoteproviderssources.

Thisexampleisatypicalmediationinstance:theuserposesqueriesintermsoftheintegratedschema(e-servicesanddomainontology)ratherthandirectlyqueryingspecificproviderinformationsources.Thisenablesuserstofocusonwhattheywant,ratherthanworryingabouthowandfromwheretoobtaintheanswers.Apartfromthenaturallanguageprocessingandthewrappingsteps,whicharenofurtherconsideredinthispaper(cf.[1]formoredetails),theMKBEEMmediationreliesontworewritingstepsachievedbytheDOS:

1.Acontextualinterpretationstep:theOFisrewrittenintoafewE-ServicesFormula(ESF)whichhavebeendynamicallyidentifiedasthecombina-tionsofe-servicesthat”bestmatch”theOF.Herearemainlyusedthee-servicesontologyandthedomainontologylayers.Inthesequel,werefertothistaskasdynamicdiscoveryofe-services.

4O.Corchoetal.

2.Aqueryplangenerationstep:theESFarerewrittenintoafewqueryplanswheretheContentProvidersAgentsthatareabletoanswertheuserqueryareidentified.Hereismainlyusedthesourcesdescriptionslayer.Inthesequel,werefertothistaskasqueryplangeneration.Theproblemofqueryplangeneration,knownasqueryrewritingusingviews,hasalreadybeenstudiedintheresearchareaofinformationintegra-tion[5,8].Thatiswhy,intheMKBEEMmediationsystem,theDOSreliesonthePicsel[8]mediatortohandlethistask.Picselisadescriptionlogicbasedinformationintegrationsystem.ItusestheALN-CARIN3languageasthecorelogicalformalismtorepresentboththedomainontologyandthecontentsofinformationsources.ThequeryrewritingalgorithmimplementedinthePicselsystemisprovedtobesoundandcomplete.Onthecontrary,thedynamicdiscoveryofe-servicesisanewresearchproblem.Thisisourproposaltosolveit,thatisdetailedinsection4,afterthepresentationofthethree-layerontology.

3ArchitectureoftheMKBEEMontology

TheMKBEEMOntologiesaredefinedasaknowledgestructuretoenablesharingandreuseofinformation.TheyprovideaconsensualrepresentationoftheelectroniccommercefieldintwotypicalDomains(TourismandMailorder)allowingtheexchangesindependentlyofthelanguageoftheenduser,theservice,orthecontentprovider.Ontologiesareusedforclassifyingandindexingcatalogues,forfilteringuser’squery,forfacilitatingman-machinedialoguesbetweenusersandsoftwareagents,andforinferringinformationthatisrelevanttotheuser’srequest.

TheALN-CARINlanguagehasbeenchosenastheontologyimplemen-tationlanguageduetoitsinferentialcapabilities.Itsdescriptionlogicpart,theALNlanguage,containsthefollowingconstructors:

•conceptconjunction(󰀌),e.g.,theconceptdescriptionparent󰀌malede-notestheclassoffathers(i.e.,maleparents),

•theuniversalrolequantification(∀RC),e.g.,thedescription∀childmaledenotesthesetofindividualswhosechildrenareallmale,

•thenumberrestrictionconstructors≥nRand≤nR,e.g.,thedescription(≥1child)denotestheclassofparents(i.e.,individualshavingatleastonechildren),whilethedescription(≤1leader)denotestheclassofindividualsthatcannothavemorethanoneleader.

•thenegationrestrictedtoatomicconcepts(¬A),e.g.,thedescription¬maledenotestheclassofindividualswhicharenotmales(females).

3

ALN-CARINisalogicallanguagecombiningDescriptionLogicsandDatalogrules[12].

Anontology-basedmediationarchitecture5

Asallthedescriptionlogics,ALNcomesequippedwithwell-defined,set-theoreticsemantics.Furthermore,theinterestingreasoningproblemssuchassubsumptionandsatisfiabilityaredecidableforALN[7].

TheMKBEEMontologiesarestructuredinthreelayers,asshowninFigure1.

Eservices Ontology󰀀E-service level󰀀

MKBEEM Global Ontology󰀀Global and󰀀domain󰀀Ontologies󰀀

Mail order󰀀Domain ontology󰀀Tourism󰀀Sources descriptions󰀀SNCF󰀀B&B󰀀...󰀀Ellos󰀀Source level󰀀

Fig.1.KnowledgerepresentationintheMKBEEMsystem.

Thesethreelayersaredescribedbelow.

1.GlobalanddomainOntologies.

TheglobalontologydescribesthecommontermsusedinthewholeMK-BEEMplatform.ThisontologyrepresentsknowledgereusableondifferentdomainswhileeachdomainontologycontainsconceptscorrespondingtooneofthedomainsoftheMKBEEMpartners(e.g;,tourism,mailorders,etc.).Thisdecompositioninglobalanddomainmodularontologieshasallowedustoreusemanydefinitionsfromexistingontologiesandmoreeasilycreatedefinitionsofthedomainontologies.Example1.Table1providesthedefinitionsoftheconceptDateandTimeoftheglobalontology.Thetraveldomainontologycontains,amongoth-ers,theconcepttripwhichhasadepartureplace(theroledepPlace),anarrivalplace(therolearrPlace)andmayhaveatransportmean(theroletransportMean),andtheconceptaccommodationwhichislocatedinaplace(roleplacedIn)andhasastartingdate(therolestartDate).2.E-servicesontology.

AlltheoffersavailableintheMKBEEMplatformareintegratedandde-scribedinthee-servicesontology.Ane-servicecanbeseenasaprovider-independentpredefinedquerycorrespondingtoanexistingofferintheMKBEEMplatform.Example2.Accordingtoe-servicesontologygiveninTable1,theMK-BEEMplatformdeliversfouroffers:

6O.Corchoetal.

•hotel,whichallowstoconsultalistofhotels.

•apartment,whichallowstoconsultalistofapartments.

•timetable1,whichallowstoconsultatripgiventhedepartureplace,thearrivalplace,thedeparturedateandthedeparturetime.

•timetable2,whichallowstoconsultatripgiventhedepartureplace,thearrivalplace,thearrivaldateandthearrivaltime.

Itisworthnotingthatthesee-servicesarespecifiedindependentlyfromagivenprovider.

3.Sourcesdescriptions.

Atthelowerknowledgelevel,sourcesdescriptionsspecifytheproviderscompetencies,i.e.,thedescriptionofthecontentsoftheprovidersinfor-mationsources.Eachcontentprovideragentisidentifiedbyitsname(e.g.,SNCF4)andcanprovidemanyviews,eachofwhichcorrespondstoapossiblequery(offer)ontheproviderinformationsource.Example3.IntheexampleoftheMKBEEMontologiesgiveninTable1,fourprovidersaredefined:SNCF,cheapJet,ExpediaandGitesDeFrance.TherearetwopossibilitiesforqueryingtheSNCFsourcetohaveinfor-mationabouttraintrips:giveninformationaboutthedepartureplaceandthearrivalplaceandeitherthedeparturedateandthedeparturetime(viewtimetableD)orthearrivaldateandthearrivaltime(viewtimetableA).ThecheapJetsourceallowstoconsultalistofplanetripsgiventhedepartureplace,thearrivalplace,thedeparturedateandthedeparturetime(viewplaneTrip).TheExpediaproviderdeliversinforma-tionabouthotels(viewExpediaHotel)whileGitesDeFrancedeliversinfor-mationaboutapartments(viewRenting)

Thefollowingsectionintroducesthee-servicesdynamicdiscoverymechanismandillustrateshowtheproposedmodulararchitectureisusedtohandletheDOSmediationtasks(i.e.,contextualinterpretationofontologicalformulasandqueryplangeneration).

4E-servicesdiscovery

IntheMKBEEMproject,thedynamicdiscoverymechanismisusedinasso-ciationwiththePicselsystemtoachievethereasoningtasksintheDOS.ThecomplementaryrolesofthesetwocomplexlogicalreasoningconstitutesthedescriptionlogiccoreforqueryprocessingintheMKBEEMsystem.Theyareinfacttwodifferentinstancesoftheproblemofrewritingconceptsusingterminologies[4].

Inthefollowingweillustratetheinterestofe-servicesdiscoveryusingtheMKBEEMontologiesprovidedintable1.

4

SNCFisthefrenchrailwaycompany.

Anontology-basedmediationarchitecture

Table1.ExampleofMKBEEMontologies.

Globalontology

.

time=(≤1hour)󰀆(≥1hour)󰀆(∀hourinteger)󰀆(≤1minute)

󰀆(≥1minute)󰀆(∀minuteinteger).

date=(≤1day)󰀆(≥1day)󰀆(∀dayinteger)󰀆(≤1month)

󰀆(≥1month)󰀆(∀monthstring)󰀆(≤1year)󰀆(≥1year)󰀆(∀yearinteger)󰀆(≤1weekday)󰀆(≥1weekday)󰀆(∀weekdaystring).trip=(≤

(≤(≥.

accommodation=(≤

(≤

11111

7

TravelDomainontology

depPlace)󰀆(≥1depPlace)󰀆(∀depPlacestring)󰀆arrPlace)󰀆(≥1arrPlace)󰀆(∀arrPlacestring)󰀆

transportMean)󰀆(∀transportMeantransportMeanType)placedIn)󰀆(≥1placedIn)󰀆(∀placedInstring)󰀆startDate)󰀆(≥1startDate)󰀆(∀startDatedate)

E-servicesontology

.

hotel=accommodation󰀆(≤1numberOfBeds)󰀆(≥1numberOfBeds)󰀆

(∀numberOfBedsnumber)󰀆(≤1hotelCategory)󰀆(≥1hotelCategory)󰀆(∀hotelCategorystring).

apartment=accommodation󰀆(≤1numberOfRooms)󰀆(≥1numberOfRooms)󰀆

(∀numberOfRoomsnumber)(≤1apartmentCategory)󰀆(≥1apartmentCategory)󰀆(∀apartmentCategorystring).

timetable1=trip󰀆(∀depTimetime)󰀆(≥1depTime)󰀆(≤1depTime)

󰀆(≤1depDate)󰀆(≥1depDate)󰀆(∀depDatedate).

timetable2=trip󰀆(∀arrTimetime)󰀆(≥1arrTime)󰀆(≤1arrTime)

󰀆(≤1arrDate)󰀆(≥1arrDate)󰀆(∀arrDatedate)

SourcesDescriptions

ProviderViewNameViewDescriptionSNCFtimetableDtimetable1󰀆(∀transportMeantrain)SNCFtimetableAtimetable2󰀆(∀transportMeantrain)cheapJetplaneTriptimetable2󰀆(∀transportMeanplane)ExpediaExpediaHotelhotelGitesDeFranceRentingapartmentExample4.LetusassumethatagivenusersubmitstotheMKBEEMsys-temthefollowingrequestexpressedinahumanlanguage:

Q1:”I’llarriveinParisonMondayandIlookforanaccommodationwithswim-mingpool”.

ThequeryQ1isfirstprocessedbytheHLPServerandconvertedtothefol-lowingontologicalformula:

OF1:”(trip)(V8),(accommodation)(V7),(arrDate)(V8,C9),(date)(C9),

(day)(C9,15),(weekday)(C9,monday),(month)(C9,april),

(year)(C9,2002),(arrPlace)(V8,paris),(leisure)(V7,swimmingPool)”.

8O.Corchoetal.

Then,givensuchanontologicalformula,thee-servicesdiscoveryisusedbytheDOStoidentifythecorrespondingrelevantservice(s)intheontologyofe-services.Thistaskisachievedintwosteps:

1.ConvertinganontologicalformulaFintoaconceptdescriptionQF

Thistaskdependsonthestructureoftheontologicalformulaandontheexpressivepowerofthetargetlanguage.InthecontextoftheMK-BEEMproject,thecurrentontologicalformulasgeneratedbytheHLPServerhaverelativelysimplestructuresthatcanbedescribedusingthesmalldescriptionlogicFL0∪{(≥nR)}.Thislogiccontainstheconceptconjunctionconstructor(󰀌),theuniversalrolequantificationconstructor(∀R.C)andtheminimalnumberrestrictionconstructor(≥nR).Inthiscase,wecanachievethistaskbycomputingtheso-calledmostspecificconcept[7]correspondingtotheontologicalformula.Example5.TheconceptdescriptionQOF1correspondingtotheontolog-icalformulaOF1giveninthepreviousexampleis:

.

QOF1=trip󰀌accommodation󰀌(≥1arrPlace)󰀌(∀arrPlacestring)󰀌

(≥1arrDate)󰀌(∀arrDate(date󰀌(≥1day)󰀌(∀dayinteger)󰀌(≥1year)󰀌(∀yearinteger)󰀌(≥1month)󰀌

(∀monthstring)󰀌(≥1weekday)󰀌(∀weekdaystring)))󰀌(≥1leisure)󰀌(∀leisurestring)2.Selectingtherelevante-services.

Thisproblemcanbestatedasfollows:givenauserqueryQFandanontologyofe-servicesT,findadescriptionE,builtusing(some)ofthenamesdefinedinT,suchthatEcontainsasmuchaspossibleofcommoninformationwithQFandaslessaspossibleofextrainformationwithrespecttoQF.WecallsucharewritingEabestcoverofQFusingT.Therefore,ourgoalistorewriteadescriptionQFintotheclosestdescriptionexpressedasaconjunctionof(some)conceptnamesinT.AbestcoverEofaconceptQusingTisdefinedasbeinganyconjunctionofconceptnamesoccurringinTwhichsharessomecommoninformationwithQ,isconsistentwithQandminimizes,inthisorder,theextrainformationinQandnotinEandtheextrainformationinEandnotinQ.Oncethenotionofabestcoverhasbeenformallydefined,thesecondissuetobeaddressedishowtofindasetofe-servicesthatbestcoversagivenquery.Thisproblem,calledbestcoveringproblem,canbestatedasfollows:givenanontologyTandaquerydescriptionQ,findallthebestcoversofQusingT.

Moretechnicaldetailsaboutthebestcoveringproblemcanbefoundin[9,10].Tosumup,themainresultsthathavebeenreachedare:

•Thepreciseformalizationofthebestcoveringproblemintheframeworkoflanguageswherethedifferenceoperationissemanticallyunique(e.g.,thedescriptionlogicFL0∪{(≥nR)}).

Anontology-basedmediationarchitecture9

•AstudyofcomplexityshowedthatthisproblemisNP-Hard.

•Areductionofthebestcoveringproblemtotheproblemofcomputingtheminimaltransversalswithminimumcostofaweightedhypergraph.•Basedonhypergraphtheory,asoundandcompletealgorithmthatsolvesthebestcoveringproblemwasdesignedandimplemented.Example6.Continuingwiththeexample,weexpectthefollowingresulttobereturnedbytheDOS:

IdentifiedservicesRestMissedinformationSolution1timetable2,leisurearrTime.hour,arrTime.minuteapartmentnumberOfRooms,apartmentCategory

Solution2timetable2,leisurearrTime.hour,arrTime.minutehotelnumberOfBeds,hotelCategoryThesesolutionscorrespondtothecombinationsofe-servicesthatbest

matchtheontologicalformulaOF1.Foreachsolution,theDOScomputestheextrainformation(columnMissedInformation)broughtbythee-servicesbutnotcontainedintheuserquery.ThecolumnRestcontainstheextrainforma-tion(leisure)containedintheuserqueryandnotprovidedbyanye-services.Thismeansthat,intheproposedsolutionstherequirementconcerningtheleisureisnottakenintoaccount.

Tocontinuewiththeexample,assumethattheuserchoosesthefirstsolution(timetable1,apartment).Then,heisaskedtocompletethemissedinformation:thearrivaltime(hourandminutes),theapartmentcategoryandthenumberofroomsintheapartment.TheresultisaglobalqueryQ,expressedasanE-serviceFormula(ESF),thatwillbesenttothePicselsys-temtoidentifytheContentProviderAgent(CPA)thatareabletoanswertothisquery.Inourexample,theexecutionofthequeryQusingtheontologiesdepictedinTable1leadstotheuniquequeryplan:[SNCFtimetableA,GitesDeFranceRenting].

ImplementationandvalidationThee-servicesdiscoveryalgorithmhasbeenimplementedasanintegratedcomponentintheMKBEEMprototype.ThisprototypeisbuiltasasetofEnterpriseJavaBeans(EJB)componentsthatinteractwitheachother.TherearealsosomecomponentsdedicatedtotheinteractionwiththeuserinterfacebuiltwithJavaServerPages(JSPs)orServlets.Finally,someofthesecomponentshavethefunctionalityofinter-actingwiththeremote(orlocallyduplicated)databasesintheproviderin-formationsystems.TheMKBEEMprototypehasbeenvalidatedonapan-Europeanscale(FranceandFinland),withthreebasiclanguages(Finnish,EnglishandFrench)andoneoptionallanguage(Spanish),intwodistinctend-userfields:1)Businesstoconsumeron-linesales,and2)Webbasedtravel/tourismservices.

10O.Corchoetal.

5Conclusion

Inthispaper,wehavepresentedanontology-basedmediationarchitecturestructuredinthreelayers:global/domainontology,e-servicesontologyandprovidersourcesdescriptions.Inthiscontext,anewreasoningmechanism,calleddynamicdiscoveryofe-services,isproposedtoallowtheuserstoau-tomaticallydiscovertheavailablee-servicesthatbestmeettheirneeds.ThisreasoningmechanismisusedinassociationwiththePicselmediatorsystemtoeffectivelyhandlethemediationtasks.Themodularityofthisarchitecturetogetherwiththeassociatedreasoningmechanismsallowtomakethewholesystemprovider-independentandmorecapabletofacethegreatinstabilityandthelittlelifetimeofe-commerceoffersande-services.Theproposedar-chitecturehasbeensuccessfullyexperiencedinthecontextoftheMKBEEMproject.

References

1.MKBEEMwebsite:http://www.mkbeem.com.

2.DataEngineeringBulletin:SpecialIssueonInfrastructureforAdvancedE-Services.24(1),IEEEComputerSociety,2001.

3.TheVLDBJournal:SpecialIssueonE-Services.10(1),Springer-VerlagBerlinHeidelberg,2001.4.F.Baader,R.K¨usters,andR.Molitor.RewritingConceptsUsingTerminologies.InKR’2000,Colorado,USA,pages297–308,Apr.2000.

5.C.Beeri,A.Y.Levy,andM-C.Rousset.RewritingQueriesUsingViewsinDescriptionLogics.InACMPODS,NewYork,USA,pages99–108,Apr.1997.6.F.CasatiandM-C.Shan.ModelsandLanguagesforDescribingandDiscoveringE-Services.InACMSIGMOD,SantaBarbara,USA,May2001.

7.F.DoniniandA.SchaerfM.Lenzerini,D.Nardi.Reasoningindescriptionlogics.InGerhardBrewka,editor,FoundationofKnowledgeRepresentation,pages191–236.CSLI-Publications,1996.8.F.Goasdou´eandM-CRoussetV.Latt`es.TheUseofCARINLanguageandAlgorithmsforInformationIntegration:ThePICSELSystem.IJICIS,9(4):383–401,2000.

9.M.S.Hacid,A.Leger,C.Rey,andF.Toumani.Computingconceptcovers:Apreliminaryreport.InWorkshoponDescriptionLogics.Toulouse,France,Apr.2002.

10.M.S.Hacid,A.Leger,C.Rey,andF.Toumani.Dynamicdiscoveryofe-services:Adescriptionlogicsbasedapproach.InProceedingsofthe18thFrenchconferenceonadvanceddatabases(BDA),Paris,pages21–25,Oct.2002.

11.A.Y.Levy,A.Rajaraman,andJ.J.Ordille.QueryingHeterogeneousInforma-tionSourcesUsingSourceDescriptions.InVLDB’96,Mumbai(Bombay),India,pages251–262.MorganKaufmann,Sep.1996.

12.A.Y.LevyandM-C.Rousset.CombiningHornRulesandDescriptionLogicsinCARIN.ArtificialIntelligence,104(1-2):165–209,1998.

13.G.WiederholdandMichaelR.Genesereth.TheBasisforMediation.InCoopIS,Vienna,Austria,pages140–157,May1995.