Information IntegrationComputational LogicLecture 17Michael Genesereth Autumn 20022QuestionGive me a list of 15 inch aluminum skillets with nonstick coating rated at least 4 out of 5 by Consumer Reports that sell for under $30 and are currently in stock.3Data SourcesDemographicDataCompanyDataPriceSheetsVendorCatalogsConsumerReportsRatingsRetailerProductDataCurrencyConversionTablesInventoryData4Information IntegrationInformation BrokerClientClientClientSourceSourceSource5•Difficulties of Using Multiple Sources– Distribution•Network delays• Part-time operation–Platform Heterogeneity (dozens)• Differences in protocol• Differences in format (XML, HTML, spreadsheet, ODBC,…)–Conceptual Heterogeneity (thousands)• Differences in schema and vocabulary•Relative incompleteness•Life is Change–Technological innovation–New Buyers, Suppliers, Market MakersComplications“73% of business managerssurveyed said they could notaccess data in their own corporatedatabases. “Gartner Group Report6Syntactic Search EnginesGoogleDocument DocumentDocumentDocument DocumentSearch Words Document References7Too Many ResultsQuery: Who is older -- Jane or John?Search Words: John Jane olderDocument Fragments:..John is older than Jane.....John is older than Jill......Jim is older than Jane...Jill wants to know whether John is older than Jane...8Too Few ResultsQuery: Is it the case that John is older than Jane?Document fragments:..Jane is younger than John.....John is more advanced in years than Jane......John is the father of Jane...9No IntegrationQuery: Is it the case that John is older than Jane?Documents:...John is older than Jill......Jill is older than Jane...10Content vs. FormThose who will not reasonPerish in the act;Those who will not actPerish for that reason.ThosewhowillnotreasonPerishintheact;ThosewhowillnotactPerishforthatreason.Semantic View Syntactic View11Structured DataFree Form Text Easy to Use but limited capability Too Few answers, too many answers Impossible to Aggregate effectivelyStructured Data Taxonomy, Attributes, Typed Values Powerful search possible Aggregation possibleAdding tags allows machine to understand so we can search and integrate.12Potential Application AreasCorporate Logistics - Enterprise Resource Directories Personnel, locations, organizations, equipment, ordersElectronic Commerce - Integrated Product Catalogs Catalogs, inventories, product ratings, contractsHealth Care - Consolidated Patient Records Doctors, nurses, lab technicians, administrators, patientsMultidisciplinary Engineering - Concurrent Engineering Architects, engineers, construction plannersCommand and Control - Situation Assessment Commanders, intelligence, field officers, consultants13 Data Integration System - integrated access to heterogeneous data sources giving the illusion of a homogeneous data management systemSemantic Integration in InfomasterInfomasterClient ClientClientSourceSourceSource“Infomaster creates an environment that makes it easierfor information consumers to get the information theyneed to answer their questions, while making it easier forowners to publish and share their databases. “Dennis Rayer, Manager, Data Warehouse,Stanford University14Databasesname manager office phone John Jill MJH222 38086Jane Jerry Cedar12 57493Jill MJH222 Jerry 420-032 5677715FragmentationHorizontal fragmentationVertical Fragmentationname manager office phone John Jill MJH222 38086Jane Jerry Cedar12 57493name manager office phone Jill MJH222 Jerry 420-032 56777name office phone John MJH222 38086Jane Cedar12 57493Jill MJH222 Jerry 420-032 56777name manager John JillJane JerryJillJerry16Conceptual Heterogeneityname manager office phone John Jill MJH222 38086Jane Jerry Cedar12 57493Jill MJH222 Jerry 420-032 56777name employee location telephone John MJH222 7238086Jane Cedar12 7257493Jill John MJH222 Jerry Jane 420-032 7256777“The biggest problem facing anyone who wants tosearch multiple structured databases. . .is that manyorganizations use different words to describe thesame thing. “Martin Marshall, Communications Week17Relational Logic for Mappingemployee(X,Y):-manager(Y,X).name manager office phone John Jill MJH222 38086Jane Jerry Cedar12 57493Jill MJH222 Jerry 420-032 56777name employee location telephone John MJH222 7238086Jane Cedar12 7257493Jill John MJH222 Jerry Jane 420-032 725677718Rules in EpilogSafe, Horn Rules grandparent(X,Z) :- parent(X,Y), parent(Y,Z).Existential Variables parent(X,f(X,Z)) :- grandparent(X,Z).Disjunction/Classical Negation father(X,Y) | mother(X,Y) :- parent(X,Y). father(X,Y) :- parent(X,Y), ~mother(X,Y). mother(X,Y) :- parent(X,Y), ~father(X,Y).Recursion ancestor(X,Y) :- parent(X,Y). ancestor(X,Z) :- parent(X,Y), ancestor(Y,Z).19Reference SchemaSchemaSchemaSchema SchemaSchemaReference SchemaSchemaSchemaRulesRulesRulesRulesRules Rules Rules20Plug-in ArchitectureSemantic IntegratorBrowserWarehouseApplication Application XML FileDatabase HTML ODBC APIPlug-insPlug-insODBC XML APIAPIApplication21Data WarehouseCentralized Data(Moving the data)MiddlewareDistributed Data(Moving the query)Spectrum of DataManagement Strategies80%10%Selective Materialization22Internet/Intranet Physical Network Hyperlinked Databases and DocumentsClick-through BrowsingUpdate (single entry principle)Cross Category SearchNotification (smart push)Dataweb User Interface“This is one of the most usefulapplications I’ve seen on the net. “ Matt Brown, Tandem ComputersDataWeb Virtual information network Hyperlinked Concepts (one entity, one page)23Demonstration ArchitectureCorningData SourceRuleLibraryIntegratorInternalWarehouseCostco Interface Payless InterfaceCorningAgentMirroAgentRegalAgentMirroData SourceRegalData SourceCostco Buyer Payless BuyerGTW InterfaceGTW Catalog User24252627282930313233343536373839404142434445Assets LeveragedComplexity hiddenSQL Skills and Code preservedAReferenceSchemaBSQLSQLEpilog46Business to Business Net Commerce$0$200$400$600$800$1,000$1,200$1,4001998 2001 2003$BillionsSource: Forrester Research, 199947Market
View Full Document
Unlocking...