1. OMG Meeting – March 2012 November 30 th 2011

2. Requirements and test cases Preliminary meta-model

3. Associations, Association Classes & Properties Provides a unification of reified and non-reified representations Closed Vs. Open DescriptionsDescriptions of concepts can be extended in multiple models Composite Concepts Support for concepts that are composites of other concepts Composition Support for composition Constraints Support for specific and arbitrary constraints Dominant Structure & Context Ability to view the same information in different decompositions Expressions Support for general expressions Expressions about individuals, types and sets using the same predicate Ability to make statements about types & metatypes Federation Semantics Specific semantics for federations Model SemanticsA way to specify the subject of a model Multiple Names for AnythingElements should not be constrained to a single or primary name Nested Models & Graphs Models should be able to contain and make statements about other models Parameterization Should support patterns and parameterization Perspective agnostic assertions The same “fact” should be able to be stated many ways Reification Choices Conceptual Information should not differ based on reification choices Roles Should support the various interpretation of roles Scope and composition of concept definitions Concepts should be able to be defined based on other, existing, concepts Subject of a Model Ability to say what a model is “about” (world, information, system, etc) Type/Instance Support for type/instance pattern of anything Units & Quantities Support for units and quantizes – conceptual models not focused on “real”, “int” Views, Viewpoints & Notations Support for multiple views and viewpoints Requirements and test cases

4. Preliminary Meta Model

5. The model is very preliminary – just getting the ideas together and starting to be validated It is not complete It is largely structural – yet to be formalized It does not yet represent a consensus Notes

6. Concepts and Context

7. Concept Relations

11. Stakeholders View of Federated Information Application Data Open Data Commercial Data Enterprise Data Partner Data Stakeholders & Systems Using Federated Data Federated Information SIMF Federation Semantic Links

12. Combining multiple independently conceived data sources and using them together for analytics and other purposes. Example: A sales department may want to combine public, internal and external information about prospect companies as part of their CRM system Key term: Independently conceived Different data sources may use different structures, technologies, vocabularies, identifiers or theories when expressing information about the same things. Sharing information between potentially independent organizations (and their independently conceived systems). Example: U.S. Government Information Sharing Environment (ise.gov) initiative to combat terrorism and other threats to the U.S. What is Information Federation?

13. Enabling collaborative processes that may cross organizational boundaries. Example: An agency wants to outsource human resources but needs to understand how the processes, services and information of their internal department can be satisfied by an external provider. Information federation is essential. Service Oriented Architecture Mediation and Brokering Example: U.S. States provide services to access healthcare information but each State’s service is different. The federal government as well as other states need to interact. Some level of mediation is required across these independently conceived services. Information exchange and federation is the essence of SOA. What is Information Federation?

14. Point-point structural transformation of data {Lets hack a solution point} Representations : XSD, XSLT, Copybooks, Code {Which are not accessible to most stakeholders} Standardized or centralized data structures or APIs {One size fits all} Representations: XSD, SQL-DD, UML, “Master Data Management” Service/API Definitions: WSDL, Corba, SoaML Canonical data model with proprietary/structural mapping {Convert to MY WAY} Representations: E/R, UML Classes, RDFS, Code/Proprietary, Data Warehouse Web data with point-point links {Publish now, federate later} Representations: RDFS+ (SEMWEB/LOD) Conceptual or logical models (sometimes) with logical links {Abstraction} Representations: Ontologies, Rules, UML (With Extensions), SBVR Federation - State of the art

15. The conceptual pivoting approach A common and growing approach to the data problem leverages abstraction: Defining a domain focused vocabulary with integrity rules and assertions as part of a conceptual model that captures domain semantics. Federation and integration is achieved by relating various logical and physical information structures to the conceptual model Information federation and integration is achieved via a “pivot” through this conceptual semantic layer This approach is used, in part, in existing standards such as CCTS (Core Components), ISO 20022 and is currently being utilized in OMG for finance. In the majority of cases the “tool” used to represent these common semantics and links is a spreadsheet, but UML and OWL are also used. Pivoting through a conceptual model

16. Cory Casanave 240 Excel UML XML There is an actual “Person”, Cory Casanave There is a concept of this person shared in this room, right now Here is one representation of him “Person” is a shared concept, independent of data structures There may also be shared agreement that Cory is a person and some other “facts” “Cory Casanave” is a name for this person He weighs 240 LBS There are multiple data representations about Cory Casanave which may or may not agree Those representations can be grounded in concepts (semantics), assisting federation Example of “Pivoting” through a conceptual model Concept of “Cory Casanave” Concept of a “Person”. Representations

17. We understand that all these representations and data structures are not “the thing” in the world But sometimes we loose site of this A conceptual domain model is a model of the things, not the data. You may also call this an ontology. A logical information model is a logical model of the data. So what is a data or information model? A model, for a specific purpose, of data about something else, that something else and that purpose are unstated. This is the root of the problem. Things Vs. Representations of Things “This is not a pipe” René Magritte Cory’s Pipe 0.02 This isn’t either

18. It must be easy to use and understand – suited to mainstream adoption! Does not require “deep” semantics, it does require well defined concepts No assumption of the same name or “matching” representations There can be any number of shared concept “theories”, no single “ring to bind them all” SIMF will be federated with other languages, so existing models in existing languages can also be federated How automated can it be? Some pivoting will be manual – but the manually asserted information saved Some will be assisted with “human in the loop” Some will be fully automatic There may be some tool variation in the automation, but not the representation Other semantic technologies will be needed to “project” the data based on the models and provide the federated information capability, completing a semantic mediation platform. There is no 100% solution – and that is not what we need. Even a minor improvement in federation capabilities could make an impact at the level of the gross national product*. More on Pivoting * Joe Bugajski, Visa International

19. Conceptual modeling with relations to structural models is not new It is done with a variety of representations UML, OWL, RDFS, E/R, Spreadsheets, FOL Ontologies, SBVR With a variety of linking and transformation mechanisms Code, XSLT, FOL, OWL, Rules, QVT, Proprietary What seems to work now – working with what we have Conceptual UML models with extensions for linking, transformed to RDF-LOD RDFS models with rules and a bit of OWL Structured English (i.e. SBVR) representations of conceptual models A bit of structural mapping, some proprietary solutions None of these approaches seem ideal for the task and all require substantial expertise, more than is practical for mainstream adoption. But, they can inform SIMF “built for purpose” standards and tools. Semantic Federation Today

20. Linked data provides a platform for data to be ubiquitously available and linked The linking and semantics of the links is not well defined OWL has largely not proved successful for wide-scale federation of independently conceived data – it is fragile, lacks expressability and is not stakeholder friendly Linked Data (RDF) could become the delivery platform for data described using SIMF OWL and RuleML Semantics can contribute to SIMF semantics, perhaps also be used as part of an implementation XSD Data structures will be federated with the SIMF conceptual model RDF representation of SIMF models provide for a SEMWEB definition Perhaps this could become the design language for Semantically Linked Data (SDL)? SIMF and Linked Data