Y. ORIENT Integrate ontology engineering into indu

ORIENT:IntegrateOntologyEngineeringintoIndustryToolingEnvironment?LeiZhang1,YongYu1,JingLu1,ChenXiLin1,KeWeiTu1,MingChuanGuo1,ZhuoZhang1,GuoTongXie2,ZhongSu2,andYuePan21APEXDataandKnowledgeManagementLab,DepartmentofComputerScienceandEngineering,ShanghaiJiaoTongUniversity,Shanghai,200030,China.fzhanglei,yyu,robertlu,linchenxi,tkw,gmc,[e-mail]2IBMChinaResearchLabNO.7,5thAve.,ShangDi,Beijing,100085,China.fxieguot,suzhong,[e-mail]panontologyengineeringtoolthatintegratesintoexistingindustrytoolingenvironments{theEclipseplatformandtheWebSphereStudiodevelopingtoolsfamily.Thispaperdescribeshowtwoimportantissuesareaddressedduringtheproject,namelytoolintegrationandscalability.WeshowhowOrientmorphsintotheEclipseplatformandachievesUIanddatalevelintegrationwiththeEclipseplatformandothermodellingtools.WealsodescribehowweimplementedascalableRDF(S)storage,query,manipulationandinferencemechanismontopofarelationaldatabase.Inparticular,wereporttheempiricalperformanceofourRDF(S)closureinferencealgorithmonaDB2database.1IntroductionOntologyisthebackbonethatprovidssemanticstotheSemanticWeb.Thede-velopmentoftheSemanticWebthusdependsheavilyontheengineeringofhighqualityontologies.NumeroustoolssupportingthistaskhavebeendevelopedinthepastsuchasProt¶eg¶e[1],OilEd[2]andWebODE[3].BasedontheR&Dexperiencesandresultsofthesetools,weinitiatedtheOrient(Ontologyen-gineeRIngENvrionmenT)project3thatdevelopsanontologyengineeringtoolthatintegratesintoexistingindustrytoolingenvironments{theopen-sourceEclipseplatformandtheWebSphereStudioTMdevelopingtoolsfamily.Inthispaper,wedescribehowtwoimportantissuesareaddressedduringtheproject,namelytoolintegrationandscalability.Theimportanceoftoolintegra-tioncomesfromthreerequirementsonthetoolandcannotbeunderestimated.?Opinionsexpressedandclaimsmadeinthispaperareoftheauthors.TheydonotrepresenttheopinionsandclaimsoftheInternationalBusinessMachineCorporation.3://®erenttoolsfordi®erentsub-tasksofontologyengineeringsuchasbuilding,mapping,mergingandevaluation.Second,becauseapplyingontologiesinrealapplicationsrequiresthetoolbeusedtogetherwithothertoolstoaccomplishacertaintask,toolintegrationisthusrequiredtoreducethecostofswitchingbetweendi®erenttools.Finally,ifthetoolcanbeintegratedwithexisitingtoolingenvironmente.g.theEclipseplatformandtheWebSphereDevelopingtoolsfamily,itwillbemuchmoredeveloper-friendlyandeasiertouse.ThesuccessstoryoftheProt¶eg¶etoolontheintegrationofnumerouspluginsalsoremindusontheadvantagesoftoolintegration.IntheOrientproject,insteadofdesigningthetoolasastand-aloneapplication,wedecidedtobuilditasasetofEclipsepluginsthatmakeitmorphintotheEclipseplatform.Thisstrategyprovestobee®ectiveinsatisfy-ingtheaboverequirementsandachievesbothdataandUIlevelintegrationwithothertools.Insection5,anexampleshowinghowOrientcanbeintegratedwithanUMLeditingtoolispresented.Futhermore,becauseWebSphereStudiodevelopingtoolsarealsobuiltontheEclipseplatform,thisstrategypavesadirectwaytotheproductfamily.Inadditiontotoolintegration,scalabilityisanotherimportantissueinOri-ent.Sinceontologies4inrealapplicationstendtohavealargesize,managinglargevolumeontologyinindustryapplicationsisthusnotafeaturebutaneces-sity.Inthispaper,wedescribehowweimplementedascalableRDF(S)storage,query,manipulationandinferencemechanizmontopofarelationaldatabaseandreporttheempiricalperformanceofthisimplementation.Inparticular,wedescribehowanalgorithmiscarefullydesignedtocalculateRDF(S)closureonarelationaldatabaseandgivetheexperimentresultofitsperformancetest.Toourbestknowledge,wedonotknowanypreviousreportonRDF(S)closureinferenceperformanceondatabasesyet,althoughtherearereportsonRDF(S)storageandqueryperformanceondatabasese.g.RQL[4]andJena2[5].Therestofthepaperisorganizedasfollows.Section2introducestheEclipseplatformandsection3describestheEclipse-basedOrientarchitecture.WethendescribetheintegrationwithEclipseUIinsection4andtheintegrationwithothermodellingtoolsinsection5.TheRDF(S)storage,query,manipulationandinferenceimplementationandempiricalperformanceanalysisispresentedinsection6.Weconcludethispaperinsection7.2EclipseToolsIntegrationPlatformAsstatedinitstechnicaloverview[6],Eclipse()isanopenextensibleIDEforanythingandnothinginparticular.Itprovidesauni-¯edmechanismtoexposeextensionpointsanddeveloppluginsthatleadstoseamlessly-integratedtools.TheEclipsearchitectureisdesignedaroundthecon-ceptsofplugins,extensionpointsandextensions.Newtoolsandfunctionalities4Inthispaper,thetermontologyreferstoaknowledgebasethatincludesconcepts,relations,instancesandinstancerelationsthattogethermodeladomaincanbeintegratedintotheplatformasplugins.Throughthispluginarchitecture,Eclipseplatformthusprovidesasolidfoundationfortoolsintegration.Besidesitswell-designedpluginarchitecture,theEclipseplatformalsoo®ersanextensibleuserint