1. The Model-Driven Semantic Web
Emerging Technologies & Implementation Strategies
Elisa Kendall
Sandpiper Software
September 8, 2005

2. Model Driven Architecture® (MDA®)
Insulates business applications from technology evolution, for
Increased portability and platform independence
Cross-platform interoperability
Domain-relevant specificity
Consists of standards and best practices across a range of software engineering disciplines
The Unified Modeling Language (UML®)
The Meta-Object Facility (MOF™)
The Common Warehouse Metamodel (CWM™)
MOF defines the metadata architecture for MDA
Database schema, UML and ER models, business and manufacturing process models, business rules, API definitions, configuration and deployment descriptors, etc.
Supports automation of physical management and integration of enterprise metadata
MOF models of metadata are called metamodels

3. MOF-Based Metadata Management
MOF tools use metamodels to generate code that manages metadata, as XML documents, CORBA objects, Java objects
Generated code includes access mechanisms, APIs to
Read and manipulate
Serialize/transform
Abstract the details based on access patterns
Related standards:
XML Metadata Interchange (XMI®)
CORBA Metadata Interface (CMI)
Java Metadata Interface (JMI)
Metamodels are defined for
Relational and hierarchical database modeling
Online analytical processing (OLAP)
Business process definition, business rules specification
XML, UML, and CORBA IDL
Model 1
Model 2
Metamodel A
Transformation Model
Metamodel B
language used
transformation
source language
target language

4. The Semantic Web
"The Semantic Web is an extension of the current web in which information is given well-defined meaning, better enabling computers and people to work in cooperation."
-- Tim Berners-Lee
“the wedding cake”
“the bus”

5. Level Setting relevant to a particular domain or area of interest.
An ontology specifies a rich description of the
Terminology, concepts, nomenclature
Properties explicitly defining concepts
Relations among concepts (hierarchical and lattice)
Rules distinguishing concepts, refining definitions and relations (constraints, restrictions, regular expressions)
relevant to a particular domain or area of interest.

6. Ontology-Based Technologies
Ontologies provide a common vocabulary and definition of rules for use by independently developed resources, processes, services
Agreements among companies, organizations sharing common services can be made with regard to their usage and the meaning of relevant concepts can be expressed unambiguously
By composing component ontologies, mapping ontologies to one another and mediating terminology among participating resources and services, independently developed systems, agents and services can work together to share information and processes consistently, accurately, and completely.
Ontologies also facilitate conversations among agents to collect, process, fuse, and exchange information.
Improve search accuracy by enabling contextual search using concept definitions and relations among them instead of/in addition to statistical relevance of keywords.

7. Semantic Web Evolution
Draft specifications for RDF/S & OWL became formal W3C Recommendations in February 2004
Increasingly attention is on applications and deployment strategies, moving from research to early adoption
Research from the DAML community and W3C is now focused on
Methodology & best practices – representing value spaces, complex relations, engineering methods, applications with richly specified ontologies
Query and rule languages
Tools and development resources
Parsers and Validators
Authoring Tools for ontologies and mark-up
Ontologies, knowledge bases, and libraries
Interaction with SOAP/WSDL to support Semantic Web Services (OWL-S, SWRL)

8. Guiding Principles
Historically, knowledge representation and reasoning systems have operated under closed world assumptions
Uncertainty is magnified under open-world, “wild, wild web” conditions, making reasoning more difficult
Semantic web languages are designed to support less certainty, provide “better” search results, informed answers to questions, not absolutes
Based on XML, they can assist businesses in leveraging mark-up, content, and data
To augment business intelligence/analysis and knowledge mining
To support knowledge sharing and collaboration, augment enterprise information integration
Enrich web services and other applications
Support policy-based applications and ensure compliance with policy
at a lower cost with higher potential ROI than traditional computing methods

9. Cost of Knowledge Capture
Most tools are oriented towards taxonomy or limited ontology formation vs. knowledge base development, terminology reconciliation, or alignment
They require significant knowledge of formal logic and domain analysis from an ontological perspective
Few are graphical and/or standards based
Vulcan estimates the cost of encoding 50 pages of basic high school chemistry textbook knowledge at $10,000 per page*
Tools to assist subject matter experts (SMEs) in encoding knowledge, including increasing automation in developing the starting points for ontology and KB development are essential
Combining MOF® and MDA® technologies with KR dramatically reduces cost, increases likelihood of success, increases availability to a much broader community of potential users
* See Vulcan, Inc. – Project Halo, at

10. MDA from the KR Perspective
EII solutions rely on strict adherence to agreements based on common information models that take weeks or months to build
Modifications to the interchange agreements are costly and time consuming
Today, the analysis and reasoning required to align multiple parties’ information models has to be done by people
Machines display only syntactic information models and informal text describing the semantics of the models
Without formal semantics, machines cannot aid the alignment process
Translations from each party’s syntactic format to the agreed-upon common format have to be hand-coded by programmers
MOF® and MDA® provide the basis for automating the syntactic transformations

11. MOF and KR Together
MOF technology streamlines the mechanics of managing models as XML documents, Java objects, CORBA objects
Knowledge Representation supports reasoning about resources
Supports semantic alignment among differing vocabularies and nomenclatures
Enables consistency checking and model validation, business rule analysis
Allows us to ask questions over multiple resources that we could not answer previously
Enables policy-driven applications to leverage existing knowledge and policies to solve business problems
Detect inconsistent financial transactions
Support business policy enforcement
Facilitate next generation network management and security applications
while integrating with existing RDBMS and OLAP data stores
MOF provides no help with reasoning
KR is not focused on the mechanics of managing models or metadata
Complementary technologies – despite some overlap

12. Metadata Management Scenarios

13. Classifying Ontologies
Level of Complexity
Level of Expressivity
Simple Taxonomy
Glossary
Topic Map
Concept Map
Hierarchical Taxonomy
Entity – Relationship
Model
Database Schema
OO Software Model
KR System
XML Schema
Classification techniques are as diverse
as conceptual models; and generally
include understanding
Methodology
Target Usage
Level of Expressivity
Level of Complexity
Reliability / Level of Authoritativeness
Relevance
Amount of Automation
Metrics Captured and/or Available

14. Model Dynamics
Model centric perspectives characterize the ontologies themselves and are concerned with their structure, formalism and dynamics.
Perspective
Level of Authoritativeness
Source of Structure
Degree of Formality
Model Dynamics
Instance Dynamics
One Extreme
Least authoritative, broader shallowly defined ontologies
Passive (Transcendent) - Structure originates outside the system
Informal or primarily taxonomic
Read-only, ontologies are static
Read-only, resource instances are static
Other Extreme
Most authoritative, narrower, more deeply defined ontologies
Active (Immanent) - Structure emerges from data or behavior
Formal, having rigorously defined types, relations, and theories or axioms
Volatile, ontologies are fluid and changing
Volatile, resource instances change continuously

15. Application Characteristics
Application centric perspectives are concerned with how applications use and manipulate ontologies.
Perspective
Control/Degree of Manageability
Application Changeability
Coupling
Integration Focus
Lifecycle Usage
One Extreme
Externally focused, public (little or no control)
Static (with periodic updates)
Loosely-coupled
Information integration
Design Time
Other Extreme
Internally focused, private (full control)
Dynamic
Tightly-coupled
Application integration
Run Time

16. Knowledge Representation vs. UML
Ontologies have well-defined, formal semantics:
To support reasoning
To represent shared semantics for intelligent agents or semantic web services
To represent declarative policies (with legal implications, such as regulatory policies)
Elements frequently cross meta-levels or apply at multiple levels (e.g., act as a class and as an individual at the same time)
Reuse of complex relationships and axioms is critical to successful development:
Relations are first class elements
Associations in UML2 are insufficient to meet this need
Individuals may be specified independently and may have properties that are not properties of the classes of which they are members (e.g., owl:sameAs, owl:differentFrom)

17. Design Rationale
Forward & reverse engineering of source material and ontologies
Support ontologies expressed in existing description logics (e.g., OWL DL) and a limited set of higher order languages (e.g., OWL Full, Common Logic)
UML Profiles vs. Metamodels
Both are required due to structural and semantic distinctions between KR languages and UML
Metamodels precisely represent the abstract syntax
Profiles facilitate use of UML notation and “code” generation
Extending the UML2 metamodel vs. new, distinct metamodels
Extension would introduce unnecessary complexity, unintended semantics
Should there be a core ODM metamodel that the others extend?
No obvious commonalities
Goal was to be true to the abstract syntax and formal semantics of the selected KR languages – without unnecessary complexity

18. Towards a Model Driven Semantic Web – Ontology Definition Metamodel
Six EMOF platform independent metamodels, (PIMs), five normative
Mappings (MOF QVT Relations Language planned)
UML2 Profiles
RDFS & OWL TM
Collateral
XMI
Java APIs
Proof-of-concepts
Conformance
All else optional
Ontology Definition Metamodel
Completed/Included Mapping
Planned Mapping
(nonnormative)

19. Impedance Mismatches & Issues
Subclassing vs. Mapping
OWL metamodel depends on RDFS metamodels
All others are related through mapping (non-normative)
Naming
Namespaces are defined at the package level in UML & MOF
Namespaces are global in RDFS/OWL, CL, & TM
Associations
Relations and Properties are “first class entities” in RDFS/OWL & CL
AssociationEnds must be defined in UML (requires model library)
Inheritance of features, attributes (slots), axioms, other relevant production rules on associations is poorly supported in UML2, not at all previously (requires use of UML2 properties, association classes)
Packaging
Discussions underway to resolve RDF/S packaging structure
RDF Schema & RDF Syntax, or possibly 3 packages
OCL is insufficiently rich to support rules, axioms

20. ODM Profiles Provide standard UML notation for each of the languages
RDFS/OWL was highest priority based on RFP
Topic Maps was high priority based on user community input
What about ER and CL?
ER has an existing graphical notation, though no “standard” notation – a profile will likely be done under a broader data modeling standardization effort (forthcoming – CWM2)
CL was originally included for constraint/rule representation; profile requested due to its utility for BSBR and interoperability with other ISO efforts; will follow through an RFC/RFP process
Essentially a one-to-one mapping from concrete meta-model element to stereotype
Abstract meta-classes contribute to base classes for stereotypes

21. ODM Mappings
OWL Full was selected as the core ODM metamodel to limit NxN mapping requirements
Final specification will include
Two-way mappings from UML2, TM, & ER to RDFS/OWL
One-way mapping from RDFS/OWL to CL
Reverse (lossy) mapping from CL to RDFS/OWL may be done through RFP/RFC
Two-way mapping from UML2 to CL (lossy) may be done through RFP/RFC
Current mapping representations include tables, abstract syntax, an interim QVT notation (i.e., inconsistent)
Plans include migration to emerging MOF Query / View / Transformation (QVT) Relations Language

22. Bridging KR and MDA

23. Technology Architecture

24. ODM Status Several revision cycles on the specification to date
Informative discussions of usage scenarios, differences between UML & OWL
Platform Independent Metamodels (PIMs) include
Resource Description Framework and Web Ontology Language (covers abstract syntax, common concrete syntactic elements from both)
Common Logic (CL), based on draft ISO CD 24707
Topic Maps (TM), based on draft ISO FCD specification
ER – based on de facto industry standards
DL Core – high-level, relatively unconstrained Description Logics based metamodel (non-normative, informational)
Revised submission (next iteration) will be posted by 11/14/05 to the OMG web site
Presentation on 8/22 revision planned for OMG Atlanta meeting (September – see
Plans for recommendation / vote for adoption December meeting

25. Implementation Strategies
Native OWL Tool
UML Tool with ODM Plug-In
RDFS/OWL
XMI Representation Enables Use in Eclipse/EMF, MOF-Based Tools
Use Reasoning to
Find Inconsistencies
Validate Logic, Policies
Use UML to Refine Linkage Ontologies
Ontology Analysis
& Validation
Java Theorem Prover
(JTP)
Hybrid Reasoning System
Link ODM Models (conceptual / semantic) To logical (ERWIN, Rose) and physical models (ERWIN, Rose, Power Designer)
MOF/ODM –
OWL Bridge
MOF Repository
Tools
Use Links / Mappings to Find & Eliminate Redundancies thru Reasoning
Use Links / Mappings For Query Planning, Business Intelligence

26. Related OMG Standards Work
Business Semantics for Business Rules
Emerging standard, Semantics for Business Vocabularies & Rules (SBVR)
Depends on the ODM CL Metamodel for logical grounding
Production Rules Representation
Emerging standard, preliminary presentation in Atlanta 9/12
Discussions in progress to relate PRR to CL
CL will provide the logical foundation for all OMG ontology and rules related standards
Mappings between ODM & SBVR, ODM & PRR, ODM & EXPRESS are planned
Relationship to emerging W3C standards for rules, semantic web services will depend on how the standards evolve
Plan in Ontology PSIG is to extend or build on ODM to support both rules and semantic web services vocabularies, potentially other vocabularies as they are developed

27. OMG Standards & Zachman Framework
Production Rules RFP
Semantics for Business Vocabularies
& Rules
Ontology Definition Metamodel

28. Related ISO Metadata Standards
ISO/IEC – Metadata Registries
Part 2: Classification for Administered Items
Part 3: Registry Metamodel & Basic Attributes
ISO/IEC – Framework for Metamodel Interoperability
Part 3: Metamodel Framework for Ontology Registration (depends on ODM Metamodels, ISO classification)
Part 4: Metamodel Framework for Mapping (depends on MOF QVT specification)
XMDR Program – Extended Metadata Registration
OWL ontologies for ISO Part 3, related MMF elements
discussions bridging the standards are underway
Liaison meeting (ODM, ISO, XMDR) planned for 9/22-23 in Toronto prior to ISO WG meeting

29. Business Integration
Semantic Web Services standards are converging (OWL-S and SWSL)
OMG RFP forthcoming for extensions to ODM to support Semantic Web Services, EXPRESS, eventually SWRL (when a rule language is selected/formalized)
Business Semantics for Business Rules joint revised submission, called “Semantics for Business Vocabularies & Rules (SBVR)” is logically grounded in Common Logic / ODM CL Metamodel
Potential mapping to forthcoming Production Rule specification
Leverage mapping from UML for BPEL to ODM extensions
Strategy:
Link business process models through MOF environment
Generate OWL for the linkage
Use linkage as basis for mediating business process semantics

30. A Framework for Next Generation Interoperability
MOF’s model management facilities and KR capabilities for machine interpretable semantics and reasoning are separate, complementary concerns
The ability of reasoners to find discrepancies in invariant rules, preconditions, and post conditions, can add scalability to MDA’s use of Design-by-Contract (DBC)
UML profiles can serve as graphical notations for Semantic Web languages, dramatically increasing ease of use
The combination of MDA and SW technologies promises to
Address the missing link in business process automation
Enable true information interoperability and continuity
Support next generation policy-based applications development

31. The Model-Driven Semantic Web
Knowledge acquisition, developing the semantics is the bottleneck
Leveraging existing assets breaks that bottleneck
Correlation through reasoning provides the utility
Multi-dimensional, cross organizational tailored semantic views
“Virtual” repository approach enables elimination of redundancy
Reasoning supports quality initiatives through inconsistency discovery, model and content validation
MDA and MOF coupled with Semantic Web technologies are the key

32. Questions / Discussion

33. John Sowa’s Top-Level Categories

34. Physical Quantities & Dimensions

35. Constant Quantities & Units

36. Standard International Units

37. Example Measurement Class Definition

38. Example Corrosion Rate Definition