The Semantic Web for Enterprise Information Architecture

							The Semantic Web for
Enterprise Information
Architecture
James Lapalme
Me, Myself and I
   Working on semantics
    and modeling problems
    since 2001 (E-learning,
    SoC)
   Enterprise architect at
    PSP Investments with a
    focus on Information
   PhD candidate at UdeM
    (MPSoC)
Objectives
   A Quick Introduction
    to the Semantic Web
   Application to
    Enterprise Data
    Architecture
   Challenges
Agenda
   Enterprise Goals and
    Challenges
   The Semantic Web
   Information Modeling
   The Semantic Web in
    the Context of EA
   Future Applications
Scary Words
   Semantics
   Ontology
   Meaning
   Conceptualization
   Model
   Formal
   Metadata
Information Challenges
   Ambiguous Semantics
       Communication
   Multiple Technologies
       Consistency
   Partially Known Value-
    Chain
       Operations
   Low Data Quality
       Decisions
   Poor Data Specification
       Expectations
The Cost of Bad Data
The Cost of Bad Data
The Roots of The Problem
Semiotic Triangle

          lize
              s   Reference




                               Re
                               fe
         bo




                                str
          m
       Sy




                                    o
      Symbol                        Referent

                  Stands for
It’s all about communication
   Effective Communication
       Common Vocabulary
       Common Understanding
       Common Context
   Examples of communication
    problems
       Mars Orbiter because of metric
        unit mismatch
       Between 1979 and 1997, 30
        changes were made to the way
        unemployment is measured
        (UK)
Semantic Web
The Web
   Created for Document Sharing
   Focused on Presentation
   Adapted for Human to Human
The Semantic Web
   Scientific American (2001)
   Focused On
     Meaning
     Knowledge Representation
     Machine Consumption
     Metadata
   « Anybody can say
    Anything about Anything
    Anywhere »
Syntax vs Semantic
   HTML and XML are
    syntax
   Machine cannot
    extract “meaning”
    from the current Web.
Just Little Theory
Set theory
Ressource Description Framework
   URIs are Surrogates for Things
   Simple Statements
       Subject, Verb, Object (triple)
   Literals based in XSD types
   Type is a standard Verb
       Uri and meaning
   XML and N3 are sterilization
RDF Schema
   Permits Information Schema
    Definitions
       Based on Set Theory and
        First-Order Logic
   Adds Subjects/Objects
       Resource, Class, Property
   Adds Verbs
       SubClassOf, Domain, Range
   Defines Entailment Rules
Web Ontology Language (OWL)
   Allows
       Schema Definitions (Description
        Logic)
       Information Schema Alignment
   Adds Subjects/Objects
       Restriction
   Adds Verbs
       subProperty, Inverse, Transitive,
        etc.
   Defines Entailment Rules
   OWL Lite, DL and Full
Semantic Web Stack
Information Modeling
Meaning
   Natural Language is
    Ambiguous
   Ambiguity can
    eliminated with
    Contextualization
   Contextualization can
    be define through
    Relations
Perception
   One Reality, Multiple
    Views of It
   Meaning is Relative to
    a Perception
   Perception is
    Contextualization
Glossary vs Taxonomy
Ontology
   Hyper-taxonomy
             intersecting
     Multiple
      taxonomies
   Meaning is define
    with rich and complex
    relations
Typical Approaches
     (Extended) Entity-Relation Modeling
              Does supports sub-classing
              Does support composite keys
              Doesn’t support restriction
              Doesn’t have a standard exchange format
               (multiple notations)
     Unified Modeling Language
              Does support sub-classing
              Does have a standard exchange format
              Doesn’t support composite keys
              Doesn’t support restriction
     XML Schemas
              Does supports sub-classing
              Does support composite keys
              Does support restriction
Why use an ontology
   Has formal semantics
       Not ambiguous
       Only one possible
        interpretation
   Based on formal logic
       Deduction on the models is
        made possible
   Human and system
    readable
       Systems can “understand”
   De facto tool for
    knowledge modeling
Applications to Enterprise
Information Management
Enterprise Information Management:
Convergence of data governance, master data
management and ontological data modeling




                      Governs       Governance




         Business                             Controls
       Information
          Model




                     Defines       Master Data
Modeling Language
   OWL is a (quasi) superset
    of traditional model
    languages
   Non-Propriety file format
   Offer Formal Verification
   Offer Test-Driven
    Development
   Analysis (SPARQL)
Model-Driven Data Specification
   Definitions (Glossary)
     Natural   Language
   Ontology (OWL)
     Relation   and Context
   Rules
     Expectation
   Alignment (CWM)
     Mapping
Ontological modeling of PSP BIM Entities




   Ontologies are used to
 define domain entities and
 the relationships between
      using W3C OWL.
Knowledge modeling vs Data
          modeling




                             OWL reasoning engine automatically
                             infers relations ships allowing rapid,
                             iterative model validation and
                             coherence as the model grows.
Data Specification Governance
   “Medium is the
    Message”
     Format   is key
   Must be owned by the
    Business
Derived Artifacts
   Databases Schemas
   XSD Schemas
   OO Models
   Cleansing Rules
   Event Models
   Knowledge Domain
    Models
Modeling Challenges
   Version Management is Immature
     How   is versioning incorpate into ontologies ?
     How do you version when everything is
      interconnected (locally vs globality) ?
   Available Tools CM tools only support
    syntax-level change management.
     How do you manage graph merging and
      semantic conflict resolution ?
Challenges
:ClassA rdf:type owl:class           :ClassA rdf:type owl:class
:PropA rdf:type owl:objectproperty   :ClassB rdf:type owl:class
:PropA rdfs:range :ClassA            :ClassB owl:disjointWith :ClassA
                                     :PropA rdf:type owl:objectproperty
:ClassA rdfs:subClassOf _:Rest
                                     :PropA rdfs:domaine :ClassB
_:Rest rdf:type owl:Restriction
                                     :ClassA rdfs:subClassOf _:Rest
_:Rest owl:onProperty :PropA         _:Rest rdf:type owl:Restriction
_Rest owl:cardinality 1^^xsd:int     _:Rest owl:onProperty :PropA
                                     _Rest owl:cardinality 1^^xsd:int
Others Challenges
   Business People don’t care or understand
    formal semantic and formal modeling.
    «   No Esperanto »
   Information consumption tools (BI) do not
    support RDF
    «   Data Silos »
Future Trends
   Semantic Databases
   RDF based Enterprise
    Application Integration
   Semantic Complex-Event
    Processing
   Semantic Business
    Intelligence
   Semantic Enterprise
    Information Integration
   Enterprise Information
    Management
       Unified Model
Questions