1. From a Single Ontologically Sound Conceptual Model to Multiple Physical Schema Languages Bruce T. Bauman, U.S. DoD 1

2. Agenda Motivation Challenge OntoUML Compiling RDFS XSD SQL Summary Further Work Conclusion 2

3. Model Types Implementation Perspective System Project Enterprise Representation Neutral Product Specific ConceptualLogicalOptimized PhysicalPhysical Technology Specific Greater Detail Fewer Models Greater Complexity Reflects Semantics Greater Reuse 3

4. Model Types - Languages Implementation Perspective System Project Enterprise NeutralProduct ConceptualLogicalOptimized PhysicalPhysical Technology OntoUML OntoUML with limited encoding annotations Technology specific tooling Full use of encoding annotations 4

5. Why existing languages don’t cut it Languages pitched at analysis and design (e.g., UML, ERD) Optimized for design targeting specific technologies Don’t have a well defined semantic mapping Implementation languages (e.g., XML Schema, DDL, OWL, RDFS) Clearly they have made implementation trades. Natural Language Not precise enough. 5

6. Unified Foundational Ontology (UFO) Created by Giancarlo Guizzardi For Conceptual / Analysis modeling Ontologically Based UML class diagram notation 6

7. Our subset of OntoUML Classes > > > Associations > > > > (merelogical relations) Attributes Datatypes > >3 > > >3 7

8. Classes ConceptRigidityIdentityExistential Dependence Example >RigidUnifiedIndependentPerson >Anti-RigidUnifiedIndependentSpouse >RigidUnifiedDependentMental State >RigidUnifiedDependentMarriage >RigidDispersiveIndependentParty > Anti-RigidDispersiveIndependentCustomer 8

9. The finished product 9

10. XSD – key assignment 10

11. Association Encoding - embed Target id : string attrib : integer Source id : string attrib : integer association AB 1..* Value Value Value Value 11

12. Association Encoding - reference Target id : string attrib : integer Source id : string attrib : integer association AB 1..* FK Value FK Value 12

13. Association Encoding - link Target id : string attrib : integer Source id : string attrib : integer association AB 1..* 13

14. Association Encoding - parts Target id : string attrib : integer Source id : string attrib : integer association AB 1..* FK Value FK Value 14

15. Association Encoding - parts Target id : string attrib : integer Source id : string attrib : integer association AB 1..* FK Value FK Value 15

16. Association Encoding - parts Target id : string attrib : integer Source id : string attrib : integer association AB 1..* FK Value FK Value 16

17. Successes / Observations It actually does work Used on many projects Model reuse is occurring Encoding rules are sufficient however new encoding patterns are still being discovered Projects select very different encodings 17

18. Challenges Hard but not too hard Physical Model Duality (encoding can seem like magic) Tools Really Hard Talent Culture Things to consider Return on Investment 18

19. Future Work Better visualization Change management Continue to improve documentation Prototype other implementations Explore semi-automatic translation 19

20. Questions Bruce Bauman Senior System Architect U.S. Department of Defense btbauma@gmail.com 20