A Framework for Knowledge Sharing and Integrity Checking for Multi-Perspective Models

1. A Framework for Knowledge Sharing and Integrity Checking for Multi-Perspective Models Yun-Heh (Jessica) Chen-Burger Artificial Intelligence Application Institute June 2001 Jessica Chen-Burger

2. Multi-Perspective Models are Used • Related Work • Zachman’s Framework • UML Modelling Suite • Ulrich Frank Group’s MEMO • Air Operation Enterprise Models • MPM is inevitable ? Jessica Chen-Burger

3. Problems for Multi-Perspective Modelling • Complexities within a single model and between models: • Problems with understanding the model • Inconsistencies within a single model and between models: • Problems with obtaining the correctness and consistency for all models • Problems with managing and reflecting (frequent) changes in the described domain Jessica Chen-Burger

4. Solution • User friendly interface: • Multi-view user interfaces (intelligent or not) • Semantic-linked traversing and browsing • Simulation, animation and abstraction of dynamic behaviours • Tutoring in model construction • Automatic V&V within a model and between models • Semantic-linked communication between models Jessica Chen-Burger

5. Approach:Using a Light-Weight Ontology (LWO) for Communication • Hierarchical and typed structure • Multiple-parents allowed • Non-circular type specifications • Knowledge is shared through the underlying common ontology Jessica Chen-Burger

6. Model-3 Model-1 Model-2 Ontology Knowledge Sharing via Light Weight Ontology Jessica Chen-Burger

7. Multi-model Communication • Describe the same problem domain • Describe different aspects of the domain • Commonly shared knowledge between models • Similar modelling principles • Pair-wise model mapping is possible • i.e. domain-model and another model • Global model mapping to some extend Jessica Chen-Burger

8. Example Mapping of Model Primitives in Different Modelling Languages • Domain Model: the light-weight ontology • BSDM: Business System Development Method • RACD: Role Activity and Communication Diagram • IDEF0 • IDEF3 Jessica Chen-Burger

9. Domain-Model BSDM RACD IDEF0 IDEF3 Example Instances Concrete Class Entity Data Control Process: action Plan, Guidelines Concrete Class Entity Role Mechanism Process: action Personnel, Equipments Concrete Class Entity Data Input Process: action Information Concrete Class Entity Data Output Process: action Information Concrete Class Process Process Function UOB Actions Jessica Chen-Burger

10. Achieving Global Consistency (1) • Local Consistency • Local consistency within each model • Pair-wise Consistency • Pair-wise consistency with domain-model • Global Consistency • Global consistency Jessica Chen-Burger

11. Achieving Global Consistency (2) • Achieve local consistency for all models • Select one model to achieve a pair-wise consistency with the domain-model to form an initial consistent set • Knowledge transfer to Domain-Model • For each discrepancy, do recursive and dependency-directed modification and convergence • Add a newmodel to the consistent set • Knowledge transfer to Domain-Model • For each discrepancy, do recursive and dependency-directed mending in the previous models and the new model • Repeat step 3 until all models are consistent with each other Jessica Chen-Burger

12. Example Rule (1)Consistent Representation of Information Jessica Chen-Burger

13. Example Rule (2) Correct Specialisation of Concepts Jessica Chen-Burger

14. Example Rule (3) Transferable Property of Full Equivalence Jessica Chen-Burger

15. Illustration Example for Rule (3) Jessica Chen-Burger

16. Summary • Current work is not completed, and will be extended and deepen in areas where appropriate; • A more rigorous approach may be established and adapted for measuring the various qualities of the rules and models using them in several aspects, e.g. • Which types of models are most suitable for those checking rules; • To which extent can such rules ensure the quality of the checked models. Jessica Chen-Burger

17. Overall Challenges • Not all modelling methods are compatible • Different level of abstraction, e.g. IDEF0 is decomposable, whereas BSDM is not • Difficulties in mapping model primitives • Not all relationships or constraints are identified • Some inconsistencies may be over-looked when multi-updating is carried out • Difficult to get an accurate global picture • The recursive mending process is exponential and human expert’s judgement must be exercised, when this occurs Jessica Chen-Burger

18. Future Work • Enhance concept mapping, i.e. to map concepts that are not “fully equivalent” but only partially equivalent • Enhance level of knowledge sharing by providing checking and conflict resolving advisory mechanism • Extend current V+V facilities by including a larger and more compete set of verification rules • Establish formal mechanism (theory) to help ensure the quality of built enterprise models • Establish measurable criteria for evaluating the quality of models • Provide a basis for assisting the process knowledge argumentation – which is the process of building Enterprise Models • Provide a basis for building workflow systems Jessica Chen-Burger

19. End of Slides Thank you for listening ! Jessica Chen-Burger

20. Advantages of using a Light-Weight Ontology • Intuitive: • Visual presentation • Hierarchical classification • Direct mapping to underlying formal representation • Concise, precise and rich in semantic: • Provides a common languages among models • Can provide a basis for semantic-related translation, integration and communication automation • Automatic V&V between models • Semantic-Linked traversing and browsing Jessica Chen-Burger

21. Challenges - Problems in Constructing the Ontology • The scope of the ontology • Level of abstraction that are captured in the ontology • What has to be said in each concept? • Classification of concepts • Naming the concepts that are suitable across different models Jessica Chen-Burger