a Semantic Web Representation

1UppsalaMaster’sThesisinComputerScience3122007-08-17ISSN1100-1836ArchivingRelationalDatabasesusingaSemanticWebRepresentationSantoshKumarReddyMaddulaInformationTechnologyComputerScienceDepartmentUppsalaUniversityBox337S-75105UppsalaSwedenSupervisor:SilviaStefanovaExaminer:ToreRisch2AbstractAdatabaseindependentmigrationtoolisdevelopedforlongtermarchivalofrelationaldatabases.TheapproachistorepresentbothschemaanddataofanexistingrelationaldatabaseinanRDF-Schemabasedrepresentation.RDF-Schemaisasemanticwebstandardforrepresentinganykindofdataandmeta-data.Thesystem,SARD,canautomaticallyconvertdatafromanexistingrelationaldatabaseintoanRDF-SchemabasedflatfilerepresentationthroughitsRelationalRDF-archivermodule.LaterSARDcanrecreatethearchiveddatainanotherrelationaldatabasebyaRelationalRDF-loadermodule.Standardrelationaldatabaseinterfaces,thequerylanguageSQL,andRDf_SchemaareusedtomakethemigrationDBMSindependent.Aproof-of-conceptimplementationshowstheviabilityoftheapproachforsimplerelationaldatabases.3TableofContents1Introduction2Background2.1SemanticWebandRDF2.2AmosII2.3SWARD3TheSARDsystem3.1UsageExample3.2Implementation3.2.1TheRDF-Archiver3.2.2TheRDF-Loader4ConclusionandFutureWork41IntroductionRelationaldatabasesareextensivelyusedforstoringinformation.Inmanycaseslongtimearchivalandrestoringthedatainrelationaldatabasesisofprimeimportance.EachDBMSvendorprovidestoolsforarchivalandrestorationofthevendor’srelationaldatabases.However,thesetoolsareonlystandardizedformigratingrelationaldatabetweendifferentversionsofthedatabaseofthesamevendor.Thisisproblematicwhenmigratingdataofarelationaldatabasefromonevendor'sdatabasetoanothervendor.ChangingDBMSvendorrequiresunloadingthedataintoasequentialfileandthenreloadedthefileintoanewdatabase.ThisrequiresthatthesequentialunloadformatisDBMSvendorindependentandsuchastandardforarchivingrelationaldatabasesisnotdefined.Furthermore,forlongtermstorageofrelationaldatabasesbeyondthelifetimeofaDBMSproductitisdesiredtohavethearchiveddatastoredinadatabaseindependentformat.Oneofthegoalswiththesemanticwebinitiativeistoprovideschemadescriptions,calledontologies,fordifferentkindsofwebresources.Metadataandontologiesplaymajorrolesforretrievingandcombininginformationfromdifferentsources.RDFandRDF-SchemaaretheW3Cstandardsfordescribingtheinformationusedinthesemanticweb.RDFallowsannotationofwebresources(e.g.URLs)withpropertiesandpropertyvalues.WhileRDFallowstoassociateanypropertywithanywebresources,theextendedlanguageRDF-Schema[1]isusedtodefineschemasofwebresources.WithRDF-Schemastandardizedproperties,i.e.ontologies,aredefinedfordifferentapplicationdomain.Theyaresimilartorelationaldatabaseschemas.Thepurposeofthisprojectistoimplementasystem,SARD(SemanticWebArchivalofRelationalDatabases),todemonstratesemanticwebbasedarchivingandloadingofrelationaldatabases.SARDcanarchivethedataofarelationaldatabaseintoastandardizedsequentialfileandlaterloadbackthearchiveddataintoanotherrelationaldatabase.Thefollowingisdeveloped:1.AnRDFschemarepresentationisdefinedthatcanrepresentbothschemaandcontentsofarelationaldatabase.2.AnRDF-archivermoduleisdevelopedthatgeneratestheRDF-schemarepresentationofagivenrelationaldatabase.53.AnRDF-loadermodulereloadsanarchivedrelationaldatabasetorecreatetheoriginalrelationaldatabase.TheimplementationofSARDutilizestheSWARDsystem[13].SWARDallowstheusertoviewthecontentsofarelationaldatabaseasRDF-Schema,givenspecificationofmappingsbetweenthedatabaseschemaandtheRDF-Schemaontology.TheRDF-archiverusesSWARDforextractingalldatastoredintherelationaldatabaseasaquerytoSWARD’sRDFviewoftherelationaldatabase.ThequeryextractsalldataaccordingtoaSARDontology.TheRDF-loaderusestheRDFAmossystem[2]toparseanRDF-SchemadocumentandthenapplydatabaseoperationsontheparsedRDFstatements.TheRDFloadercallsSQLstatementsthatrecreatestherelationaldatabaseschemaandpopulatesthenewdatabase.Thisprojectisproofofconceptonlyandisdevelopedforonlyverysimplerelationaldatabases.ThesystemcanarchivetableshavingonlythebasicdatatypesVARCHAR,INTEGER,DATE,andFLOAT.Otherdatabasefeaturessuchascompoundkeys,foreignkeys,triggers,constraints,advanceddatatypes,etc.arenothandled.62BackgroundSARDisbasedonthefollowingtechnologies:relationaldatabases,RDF,SWARD,andAmosII.Theconceptsofrelationaldatabasesareassumedknowntothereaderandaredescribedinmanytextbooks,e.g.[14].2.1SemanticWebandRDFInarelationaldatabasethedescriptionoftablesisprovidedbytheschema.Similarlysemanticweb[3]representationscanbeusedtodescribetheschemaofwebresources.ThesemanticwebisawayofrepresentingWorldWideWebinformation,inwhichtheinformationisexpressednotonlyforinterfacingusersaswithHTML,butalsoinaformsothatthatisunderstandableandusedbyprogrammersandprograms.RDF[4]isthebasiclanguagefordescribingwebcontentsusedinthesemanticweb.AnyexpressioninRDFisrepresentedasacollectionofRDFtriples[4],alsocalledRDFstatements.AsetofsuchtriplesiscalledanRDFgraph.AnRDFtriplecontainsthreecomponents:•Thesubjectisusedtoidentifytheentitydescribedbythestatement.Forexample,ifweconsideranRDFstatementthatstatesthatthedocu