A Web Information Retrieval System Architecture Based on Semantic MyPortal

HaiboYu1,TsunenoriMine2,andMakotoAmamiya2

DepartmentofIntelligentSystems,{GraduateSchool1,Faculty2}ofInformation

ScienceandElectricalEngineering,KyushuUniversity6-1Kasuga-koen,Kasuga,Fukuoka816-8580,JAPAN

{yu,mine,amamiya}@al.is.kyushu-u.ac.jp

Abstract.Inthispaper,wemainlyfocusonacommunicationmecha-nismwhichenablesefficientinformationpublishingandsharingamongsemanticdesktops.WeproposeMyPortalasa“onestop”foralltheinformationrelevanttotheuserandfurtherproposetheconceptualar-chitectureofaP2PcommunityWebinformationretrievalsystembasedonMyPortal.ThisarchitectureenablesnotonlypreciselocationofMy-PortalinstancesandtheirWebresourcesbutalsotheautomaticorsemi-automaticintegrationofhybridsemanticinformationdeliveredthroughWebcontentandWebservices,anditalsoensuresthatthesemanticswillnotbelostduringanypartofthelifecycleoftheinformationretrievalprocess.

1Introduction

CurrentWebdesigntargetshumanconsumption,basedonkeywordsforinfor-mationindexingandsearching,whichnotonlygivesrisetoanenormousnumberofirrelevantsearchresponses,butisunsuitableformachineprocessing.Inad-dition,theuser’sdesktopinformationandthepublishedWebinformationaremanagedseparately,givingrisenotonlytoaredundancyofinformationbutalsocreatingdifficultiesinmanagingtherelationshipamongitemsofinformationandapplyinguserpersonalization.

Currently,therearesomeresearchprojects,suchasHaystack[1]andGnowsis[3]tryingtousesemanticWebtechnologyforthemanagementofuserpersonaldesktopinformation.However,theylackthefunctionalityforsearching,access-ing,aggregatingandprocessingoftheWebinformationontheflywhennecessaryandaunifiedinterfaceformanagingnotonlythepersonaldesktopinformationbutalsotherelevantWebinformation.Andareasonablearchitectureandeffi-cientmechanismsfortheconnecting,discovering,andsharingoftheinformationamongsemanticinformationnodesarenecessary.

Inthispaper,wemakeourmainconcernonhowtoconnecttheseinformationnodesinarobustandefficientway,howtodiscoverandsharetheinformationamongtheseinformationnodesandwhatfunctionalitiesneedtobeprovidedinordertorealizethesetargets.

WeproposeoursemanticWebinformationretrievalsystemarchitecturebasedonthefollowingmainideas.

First,“combiningWebportaltechnologywithsemanticdesktoptechnologytoprovidea“onestop”fortheusertoallhisrelevantinformation.”Asseman-ticdesktopprovidesagoodsolutionformanaginguserpersonalinformationbutlacksthefunctionalitytosearch,collectandaggregateinformationfromtheWebfortheuseronthefly.Ontheotherhand,Webportalsprovideagoodsolutionforcollectingrelevantinformationfortheuser,butlackoptionsforpersonaliza-tionandsufferfromtheproblemsofcentralizedarchitecture.WemakeuseofthebasicmechanismsforsemanticpersonalinformationmanagementofcurrentsemanticdesktopsandenhancetheirWebinformationpublishingandsharingfunctionalitiestoconstructasemanticMyPortal.

Second,“usingpeer-to-peercomputingarchitecturetoconnectMyPortalswithemphasisonanefficientmethodforreducingcommunicationload.”De-centralizedP2Psystemsarerobust,scalableandcheaptomaintain,buttendtohavelargeamountsofinformationtransferredamongmanypeers.Hence,anefficientmechanismforreducingcommunicationloadswithleastlossofprecisionandrecallisveryimportantinaP2Pinformationretrievalsystem.WeproposeourAgent-Community-basedPeer-to-PeerinformationretrievalmethodcalledACP2PtoconnectandmanagethecommunicationamongMyPortals.

Third,“ensurethatthesemanticsarenotlostsightofduringanypartofthelifecycleofinformationretrieval.”Inordertoenableconsumerre-usingsemanticdata,wedesignedtheinterfacesandtheprotocolsinvolvedinthewholelifecycleofinformationretrievaltaskswithsemantictechnology.

Fourth,“allparticipantscontributetothesemanticdescriptionconsistently.”Efficientsearchingforhighqualityresultsisbasedonpertinentmatchingbe-tweenwell-definedresourcesanduserqueries,wherethematchingreflectsuserpreferences.WeuseWebsitecapabilitydescription(WSCD)todescribethecapabilitiesofMyPortalandsubmituserqueriesconsistently.

Fifth,“integratingWebinformationdeliveredthroughWebcontentsandWebservices.”ConventionalWebcontentsandWebserviceshavebeenmanagedseparatelyastheytargeteddifferentconsumer,wewillsupporttheintegratedmanagementofsemanticWebcontentsandWebservicesatdifferentlevelsinMyPortal.

2MyPortal

MyPortalisa“onestop”thatlinkstheusertoalltheinformations/heneeds.Itisattheuser’sowndesktop,whichisalsoaWebserveritselfandisdesignedtomanageuser’spersonalinformationwithsemanticWebtechnologyinaflexiblepersonalizedway.Itprovidesbothsemanticbrowserandsemanticsearchenginefunctionalitiesandthesefunctionsmanagenotonlylocaluserdesktopinforma-tionbutalsotheremotesemanticMyPortalinformation.ItsinformationcanbepublishedthroughWebcontentsandWebservicesandsharedbyotherswithproperauthority.

ThestructureofMyPortalisshowninFig1.Itconsistsoffollowingfourcomponents:corecomponentprovidesbasicsupportforsemanticWebtechnolo-giesandknowledgemanagement,userinterfacecomponentprovidesaunifiedinterfaceforcreating,browsing,querying,andmanagingoftherelevantinfor-mation,desktopinformationmanagementcomponentmanagestheconventionalpersonalinformationsuchasdocuments,e-mail,contactinformation,andcom-municationcomponentwhichisthedelegateoftheuserforcommunicationwithotherMyPortals.

Communication ComponentInformation RetrievalAgent (IRA)History ManagementAgent (HMA)Core ComponentOntologyWeb ServicesManagementKnowledge ManagementInference EngineQuery EngineTransformationUser Interface AgentUser InterfaceDesktop Information ManagementAdaptors Desktop InformationFig.1.StructureofMyPortal

Onecanreferto[4]foralittlemoredetailforMyPortal.

3

ConceptualArchitectureofWebInformationRetrievalSystemBasedonMyPortal

OurconceptualarchitectureforacommunitysemanticWebinformationretrievalsystemisillustratedinFig2.

Thearchitectureconsistsofthreemaincomponents:a“consumer”whichsearchesforWebresources,a“provider”whichholdscertainresources,andamediatorwhichenablesthecommunicationbetweentheconsumerandtheprovider.Inourarchitecture,theprovidersandconsumersareallMyPortal.EachproviderdescribesitscapabilitiesinwhatwecallaWSCD(Websiteca-pabilitydescription),andeachconsumerwillsubmitrelevantqueriesbasedonuserrequirementswhenaWebsearchisnecessary.ThemediatoriscomprisedofagentsassignedtotheconsumerandprovidersusinganAgent-Community-basedP2Pinformationretrievalmethodtofulfillthesearchandaccesstasks.3.1

ConnectingMyPortalswithACP2Pmethod

ThecommunicationbetweenconsumerandprovidersisbasedonanAgent-Community-basedPeer-to-PeerinformationretrievalmethodcalledACP2Pmethod[2],

ProvidersMyPortal1WSCD(GID, WCD, WSD)MyPortal2WSCD(GID, WCD, WSD)MyPortaln…WSCD(GID, WCD, WSD)IR AgentIR AgentIR AgentMediatorIR AgentUI AgentHM AgentMyPortalConsumerFig.2.AConceptualArchitecture

whichusesagentcommunitiestomanageandlookupinformationrelatedtoauserquery.

Inordertoretrieveinformationrelevanttoauserquery,anagentusestwohistories:aquery/retrieveddocumenthistory(Q/RDHforshort)andaquery/senderagenthistory(Q/SAHforshort).MakinguseoftheQ/SAHisexpectedtohaveacollaborativefilteringeffect,whichgraduallycreatesvirtualagentcommunities,whereagentswiththesameinterestsstaytogether.

TheACP2Pmethodemploysthreetypesofagents:userinterface(UI)agent,informationretrieval(IR)agentandhistorymanagement(HM)agent.Asetofthreeagents(UIagent,IRagent,HMagent)isassignedtoeachuser.AlthoughaUIagentandanHMagentcommunicateonlywiththeIRagentoftheiruser,anIRagentcommunicateswithotherusers’IRagentstosearchforinformationrelevanttoitsuser’squery.ApairofQ/RDHandQ/SAHhistoriesandretrievedcontentfilesaremanagedbytheHMagent.

TheACP2PmethodisimplementedwithMulti-AgentKodama(Kyushuuni-versityOpen&DistributedAutonomousMulti-Agent)[6].Kodamacompriseshierarchicalstructuredagentcommunitiesbasedonaportal-agentmodel.Apor-talagentistherepresentativeofallmemberagentsinacommunityandallowsthecommunitytobetreatedasonenormalagentoutsidethecommunity.

WearecurrentlyplanningtouseSPARQLRDFquerylanguageandSPARQLprotocolasoursemanticcommunicationinterfacesbetweenprovidersandcon-sumers.3.2

Websitecapabilitydescription(WSCD)

ResourcelocationisbasedonmatchingbetweenuserrequirementsandWebsitecapabilities,henceacapabilitydescriptionofMyPortalisnecessary.WedescribethelayeredcapabilitiesofMyPortalbylayers.

First,wesemanticallydescribethegeneralcapabilitiesoftheWebsite,andwecallthisa“generalinformationdescription(GID).”TheGIDgivesanexplicitoverviewoftheWebsitecapabilitiessuchastheircategory,topic,andcanbeusedastheinitialfilterforjudgingcongruencewithuserpreferences.Second,we

givetheWebcontentcapabilitydescription(WCD),itisthemetadataofWebcontentsandiscomposedofknowledgebasesofalldomainsinvolved.Third,wegivetheWebservicecapabilitydescription(WSD)whichisfurtherexpressedbytwolayers:“asemanticWebservicedescription(SWSD)”and“aconcreteWebservicedescription(CWSD).”Thishierarchicalcapability-describingmech-anismenablessemanticandnon-semanticWebservicecapability-describingandmatchmakingfordifferentlevels.

ForthedetailsofourWebsitecapabilitydescriptionmechanism,onecanrefertodocument[5].

4Conclusion

Inthispaper,weaddressedourmainideasonconstructingaP2Pcommunityse-manticWebinformationretrievalsystembasedonMyPortal,mainlyfocusedonhowtoconnectMyPortalstoenableautomaticandefficientinformationshar-ingandwhatfunctionalitiesarenecessarywhenconstructingaMyPortal.Inthefuture,wewillrealizeaprototypeofMyPortalandaP2PcommunityWebinformationretrievalsystembasedonMyPortal,andevaluatetheeffectivenessofourapproaches.ExperimentsinusingtheACP2PmethodforsemanticWebdataretrievalinadynamicmultiplecommunityenvironmentwillalsobecarriedout.

References

1.D.Huynh,D.Karger,andD.Quan.Haystack:APlatformforCreating,OrganizingandVisualizingInformationUsingRDF.InProceedingsoftheInternationalWork-shopontheSemanticWeb(atWWW2002),2002.http://semanticweb2002.aifb.uni-karlsruhe.de/proceedings/Research/huynh.pdf.

2.T.Mine,D.Matsuno,A.Kogo,andM.Amamiya.Designandimplementationofagentcommunitybasedpeer-to-peerinformationretrievalmethod.InProc.ofEighthInt.WorkshopCIA-2004onCooperativeInformationAgents(CIA2004),LNAI3191,pages31–46,92004.

3.L.Sauermann.TheGnowsisSemanticDesktopforInformationIntegration.InIOAWorkshopoftheVM2005Conference,2005.

4.H.Yu,T.Mine,andM.Amamiya.TowardsaSemanticMyPortal.InThe3rdInternationalSemanticWebConference(ISWC2004)PosterAbstracts,pages95–96,2004.

5.H.Yu,T.Mine,andM.Amamiya.TowardsAutomaticDiscoveryofWebPortals-SemanticDescriptionofWebPortalCapabilities-.InSemanticWebServicesandWebProcessComposition:FirstInternationalWorkshop,SWSWPC2004,LNCS3387/2005,pages124–136,2005.

6.G.Zhong,S.Amamiya,K.Takahashi,T.Mine,andM.Amamiya.TheDesignandImplementationofKODAMASystem.IEICETransactionsonInformationandSystems,E85-D(4):637–6,April,2002.