Semantic Applications for Financial Services

Semantic Applications for Financial Services

• 1.
  Semantic Applications forFinancial ServicesDavid NewmanStrategic Planning ManagerEnterprise Technology Architecture and PlanningWells Fargo BankJune 23, 2010
• 2.
  Disclaimer1Thecontent in this presentation represents only the views of the presenter and does not represent or imply acknowledged adoption by Wells Fargo Bank. Examples used within are purely hypothetical and are used for illustrative purposes only and are not intended to reflect Wells Fargo policy or intellectual property.
• 3.
  What Benefits DoesSemantic Technology Provide for Financial Services Organizations?What are some of the business and technology drivers for Semantic Technologies from a Financial Services perspective?What are some of the critical business and technology problems that Semantic Technology attempts to remedy?What are some limitations with conventional Information Technologies that Semantic Technology improves upon?What are some Financial Service use cases that can demonstrate benefit by using Semantic Technology?2
• 4.
  IT Organizations areoften asked by the Business to:provide a holistic, comprehensive, integrated view of the Customerfulfill major data and system integration initiativescross organizational and system boundaries to accomplish thisdeliver all of the above functionality faster, cheaper, smarter3Common IT Challenges at Financial Services Firms …
• 5.
  This must oftenbe accomplished in environments where there exists:a preponderance of incompatible data definitions, vocabularymultiple incompatible physical data and file formats, databases, storage mechanismsa proliferation of fragmented, redundant dataa proliferation of unstructured data that is inaccessible to most usersdissonance between the business stakeholders definition of data and processing rules and how such data and rules are actually codified within application softwareCan result in high costs, slipped dates4Often Requires IT to Surmount Difficult Obstacles …
• 6.
  Requires New andInnovative Tools that will help IT organizations to: standardize and unify the meaning of data across the enterprisecapture and persist business and technical knowledge as information assetsfoster data integration despite organizational boundariesgive greater control to the Business for definitions of data and business rulesproduce better results faster and cheaper than conventional technologiesSemantic Technology can help to achieve these goals!5That May Not Always Be Solved by Conventional Technologies
• 7.
  6Data SchemaNew DataEntityNew Physical Table for New EntityBusiness Rules in CodePhysical DatabaseApplication SoftwareAccessDefineUpdateConventional Information Technologies: What’s Wrong?Conventional Technology Data Definition and Access PatternsKnowledge is encapsulated in opaque softwareData organization is tightly coupled with the schemaData schemas reflect limited knowledgeSchemas enforce limited data integrityData is siloed as is its meaningData fragmentation
• 8.
  Data redundancy
• 9.
  Data incompatibility
• 10.
  Labor intensive tasks
• 11.
  High costs
• 12.
  Slow time tomarketConventional Relational Database ExampleAwareness of the physical organization of data is necessary
• 13.
  Many tables areoften required to capture entities and their relationships
• 14.
  Entity relationships arerealized by joining data, mainly by its keys
• 15.
  Guided by ClosedWorld Assumption – if data is not present it does not exist7
• 16.
  8Aligns linguistically withhow we think and speak!Subject(domain)Predicate (property)Object(range)RDF Triples/ StatementsWhat is Semantic Technology?Major step towards reducing data chaosBased upon Description LogicA mathematically verifiable symbolic logic that allows reasoning about entities and the many properties that describe entity relationshipsDescribes entities in terms of:Concepts (classes)Relationships (properties)Individuals (instances)Makes inferencing possibleInfers relationships and memberships in classes per axioms via a “Reasoner”Guided by Open World AssumptionIf data is not present it maystill exist!Jackson Pollock “Convergence”
• 17.
  9Some Business RulesRemoved from CodePhysical Format Unchanged after New Data Entity AddedNew Data EntityOntology / Semantic SchemaPhysical DatabaseApplication SoftwareAccessDefineUpdateSome Business Rules Added to OntologySome Inferred DataSemantic Technology: How does it Help?TBoxTBox (terminology) ABox (assertions)Semantic Technology Data Definition and Access PatternsMeaning is consistent
• 18.
  Knowledge is accessible
• 19.
  Applying rules todata is easier and less costly
• 20.
  Data access costsshould be lower
• 21.
  Faster time tomarketKnowledge is open and represented by an ontologyMeaning and relationships of data definedData organization is decoupled from the schemaInferencing creates new knowledgeAll semantic data is Web addressable
• 22.
  Financial Information Ontology10Business PartnerAccountAccount StatusaccountStatusConsumerAccountPersonClosedOpenaccountForStatusDepositAccountConsumer Credit AccountProductproductTypeChecking AccountBusinessEntityHELOCisAccountEvent TypehasAccounttitleHolderis CustomerhasIdentityCustomerOnline LoginisEligibleForConsumer ProductCredit Card Eligible CustomerAccountOpenGold Credit Card Eligible CustomerhasPre-QualifiedConsumer CreditCredit CardHomeEquityCredit Risk Retail CustomerChangeAddressFraud Risk Retail CustomerTransferFundsRetail DepositCountryonlineLoginEventLocationdescribes EventUnited StatesSavingsDeposithasEventeventForCustomerRetailCheckinghasEventTypeEventBad CountryOnline Login EventeventForCountryBad Country XSuspicious Online Login Event
• 23.
  SemanticRDF “Triple Store” Example11Every element is a Web addressable URI!TBoxTerminologyOntology SchemaInferredTriplesABoxAssertions FactsData
• 24.
  How Can WeApply Semantic Technology to Specific Financial Services Use Cases for Maximum Benefit?The following use cases represent a sampling of ways that Semantic Technology can be effectively applied in a Financial Services organization:Linked Enterprise Data: 360 Degree Customer ViewRDFa Enablement of Online Financial Services and ProductsFraud DetectionEligibility and Suitability RulesCredit Risk ManagementIntegrated Financial StatementsConcept Extraction and Categorization from Unstructured TextMarket Intelligence for Investment Analytics12
• 25.
  Linked Enterprise Data:360 Degree Customer View13Semantically enabled data that is Web addressable and “inter-linked” across the enterpriseTranscends organizational boundaries and provides universal access to data wherever it resides within the enterprise (and externally)Reduces redundancy by obtaining data from its “virtualized” sourceIntegrating data in a siloed environment is a major win for Financial Information Systems“KYC”Know YourCustomerDepositsOnlineBankingCustomer360 degree view of customerLoansStores“Customer Centric”EnterpriseData CloudCorporateTwitterFacebookRiskCredit Bureau
• 26.
  RDFa Enablement ofOnline Financial Services & ProductsFinancial Ontologies14RDFa Enabled Web PageBase OntologySemantic search engine optimization, semantic marketing and salesGrowing evidence that RDFa will:improve rank on Search Enginesincrease traffic to siteimprove click-thru ratesFIs can RDFa enable:products information , e.g. terms, rates for:loans, CDs, checking, savings, etc. services e.g. bill payments, financial adviceContext based semantic search
• 27.
  Semantic agents
• 28.
  Agent initiated search
• 29.
  Agent initiated filtering
• 30.
  Agent initiated transactionsFraudDetectionEffective Pattern Detection and Event CorrelationFrom a set of known facts (Events), and a set of rules (Axioms), a Reasoner infers membership in classes that reflects a specific pattern indicating fraud riskExpectation of reduced cost in comparison to conventional technologiesalter Tbox dynamically to define new patterns and relationshipsyields faster results, lower maintenance and deployment costsUtilizes Open World Assumption (OWA) ReasoningEffective Link AnalysisOnce a common entity is known e.g. bad phone number; other risky relationships may emerge by invoking queries that perform Graph Pattern Matching to identify fraud networks or other victimsUnique Naming Assumption not supported by OWA 15
• 31.
  Fraud Risk Pattern16Whichconsumer customers might be atrisk of Online Account Takeover Fraud? AccountCountryConsumer AccountAnswers the Query:onWatchListFor some OnlineAccountTakeoverand returns a set of customers at riskProductEquivalent Class:Account and productType some ConsumerProductBadCountryproductTypeBad Country XConsumer ProductRisk CategoryhasAccountonlineLoginEventLocationOnlineAccountTakeovernCustomerEventeventForCustomerRetail CustomerOnline Login EventSuspicious Online Login EventEquivalent Class:Customer and hasAccount some ConsumerAccountEquivalent Class:eventForCustomer some Customer and isEventType value OnlineLogin and onlineLoginEventLocation some BadCountryhasEventFraud Risk Retail CustomerEquivalent Class:RetailCustomer and hasEvent some SuspiciousOnlineLoginEventand hasEvent some (Event and isEventType value AccountOpen)and hasEvent some (Event and isEventType value ChangeAddress)isEventTypeEvent TypeOnline LoginAccountOpenChangeAddressonWatchListForNote: Semantic solutions, other than OWL DL, could also be used to achieve the same results
• 32.
  Eligibility and SuitabilityRulesOutbound marketing campaign extractions based upon pre-qualification of customers for specific productsCross-Sell and Offer GenerationOnline preferences and PersonalizationEligibility rules for:Account AcquisitionsLoan OriginationsMoney Movement Transactions17
• 33.
  Credit Card Eligibility:Which consumer customers are eligible for a Consumer Credit Card?AccountAccount StatusRetail Checking AccountaccountStatusClosedOpenbalanceEquivalent Class:Account and productType some RetailCheckingDoubleoverDraftsMosProductIntegerproductTypeConsumer ProducthasAccountConsumer CreditCustomerConsumerCredit CardBasicCreditCardCredit Card Eligible CustomerGold Credit CardisEligibleForEquivalent Class:Customer and hasAccount some (RetailCheckingAccount and accountStatus value Open and balance some double[> 1000.00] and overdraftsMos some nonNegativeInteger[< "1”])Retail DepositSavingsDeposithasPrequalifiedRetailCheckingGold Credit Card Eligible CustomerAnswers the Query:isEligibleFor some BasicCreditCardand returns a set of eligible CustomersCan also answer:isEligibleFor some GoldCreditCardand returns a set of Customers eligible for all Premium Consumer Credit Card ProductsEquivalent Class:CreditCardEligibleCustomer and hasAccount some (RetailCheckingAccount and balance some double[> 50000.00] 18
• 34.
  Credit Risk ManagementIdentifylevels of credit risk by vetting a set of facts collected about the customer with a set of rules that govern risk19
• 35.
  Which consumer customersare at risk from a credit perspective?Credit Risk Management: AccountRetail Checking AccountConsumer Credit AccountEquivalent Class:Account and productType some RetailCheckingEquivalent Class:Account and productType some ConsumerCreditProductproductTypeConsumer ProductIntegerdelinquentDaysIntegeroverDraftsPastMonthConsumer CredithasAccountCredit CardHomeEquityCustomerCredit Risk Consumer CustomerRetail DepositEquivalent Class:Customer and ((hasAccount some (RetailCheckingAccount and (overdraftsPastMonth some nonNegativeInteger[> 1]))) or (hasAccount some (ConsumerCreditAccount and (delinquentDays some nonNegativeInteger[>= 30]))))RetailCheckingSavingsDepositRisk CategoryConsumerCreditRiskatRiskForAnswers the Query:atRiskFor some ConsumerCreditRiskand returns a set of Customers at risk20
• 36.
  Integrated Financial StatementsSemanticallyaligned financial statementsability to roll up financial information across disparate internal business units and external companies so that Financial Reports can be published/understood with higher levels of reliability and trustattain holistic view of organization’s financial healthimprove financial risk management for enterprise and investmentsXBRL (Extensible Business Reporting Language)SEC has mandated US public companies file financial reports using XBRL 3Q09 (proliferating globally)RDF/OWL enabled XBRL XMLLeverage benefits of semantic technology using XBRLW3C Interest GroupRhizomik Initiative (ReDeFer  XML2RDF,XSD2OWL)21
• 37.
  Concept Extraction andCategorization from Unstructured Text eDiscovery - categorizing, searching and accessing structured and unstructured content often for legal purposesSemantic metadata tags associated with content for optimized searchIntentionally asserted by userAutomatically asserted by using semantic entity extraction and Natural Language Processing (NLP) tools to identify conceptual meaning of unstructured contentCustomer Related Concept ExtractionSemantic parsing of unstructured text using NLP and concept extraction to identify:Meaning of customer emails for Voice of the CustomerDirecting notifications to Bankers based upon accurate categorization of content from text for Lead Management purposes. etc.22
• 38.
  Market Intelligence forInvestment AnalyticsOpen Source Intelligence capabilitiesleverage semantic analysis of news feeds and Web content pertaining to companies of interest for investment purposes Low Latency Critical Notifications to Analystsprovides rapid categorization of content and processing against a set of semantically defined criteria (axioms) in order to send notifications to analysts for further investigation when an event of interest is identified23
• 39.
  SOABetter Service OrientedArchitectureService Oriented Architecture
• 40.
  SOA is oftenfoundational to Financial Service architectures
• 41.
  Canonical Semantic DataSchema can auto-translate data content from one interface protocol to another, increasing the level of interoperability
• 42.
  Semantic Service Repository
• 43.
  Elements within schemadefined as entities within an ontology ensuring semantic alignment, clarity and expressiveness
• 44.
  Service Registry andDiscovery
• 45.
  Requires capability toadvertise and locate service interfaces defined by a Service Registry using semantic content for unambiguous context based search24
• 46.
  25EAImplications for EnterpriseArchitectureand Data Management OrganizationsEnterprise Ontology Governance:
• 47.
  Manage and providestandards and quality control for enterprise semantic content across the enterprise
• 48.
  Enable and managean enterprise Ontology Repository
• 49.
  Limit risk ofsiloed ontologies, enable federation of ontologies
• 50.
  Evolve enterprise “UpperOntology”
• 51.
  Business Semantics Management
• 52.
  Encourage and incubateLine of Business Ontologies
• 53.
  Influence LOBs toopen their data silos for Linked Enterprise Data
• 54.
  Provide business friendlyinterface that front-ends the Ontology
• 55.
  Semantic Technology Governance
• 56.
  Ensure effective AccessControl and Trust mechanisms are provided
• 57.
  Ensure effective Qualityof Service mechanisms are provided to achieve desired performance, availability, recoverability (etc.) standards26Triple/RDF StoresOntology EditorsReasonersMiddleware(Some) Providers of Semantic TechnologyLanguagesOWLAPIPelletRacerProSesameNetKernelBusiness SemanticsManagement
• 58.
  27(Some) Recommended Books
• 59.
  28Follow-up or QuestionsEmail:[email protected]