COE

1
Ontology and Database Confluence
DIPData, Information, and Process Integration
with Semantic Web Services

• Stefano Spaccapietra
• Database Laboratory (LBD)
• Ecole Polytechnique Fédérale Lausanne, EPFL
• Switzerland

2
The New Driving Force

• From business-oriented IT
• Information Processing
• To IT for Individuals
• Information Exchange
• Information Services
• Personalization
• Shareable Semantics

3
DIP General Visionhttp//dip.semanticweb.org//
4
DIP Research Focus
5
DIP Partners

• SAP AG, Germany
• Tiscali Österreich GmbH, Austria
• Fundacion De La Innovacion.Bankinter, Spain
• Berlecon Research GmbH, Germany
• Essex County Council, UK
• Forschungszentrum Informatik, Germany
• inubit AG, Germany
• Net Dynamics Internet Technologies GmbH u. Co KG,
  Austria
• Sirma AI Ltd., Bulgaria
• The Open University, UK
• Unicorn Solution Ltd, Israel
• Vrije Universiteit Brussels, Belgium
• Intelligent Software Components, S.A, Spain

• National University of Ireland, Galway, Ireland
• British Telecommunications Plc., UK
• Swiss Federal Institute of Technology,
  Switzerland
• Universität Innsbruck, Austria
• ILOG SA, France

6
DIP Workpackages
7
EPFL-LBD in DIP

• Ontologies
• Design
• Context-Awareness
• Modularization
• Integration
• Distributed Reasoning

8
Ontology Definition

• An explicit specification of a conceptualization
  
• Ontology a means to share information and to
  achieve semantic interoperability between humans
  and computers
• An ontology describes a formal specification of a
  certain domain
• Shared understanding of a domain of interest
• Formal and machine manipulable model of a domain
  of interest

9
An Ontology is ...

• somewhere (could be
  centralized or distributed)
• some set of (definitely not
  necessarily a partition)
• somehow related terms (ontology
  language definition)
• whose use has to some extent been agreed upon
• preferably with some explanation of their meaning
• Ontologies are also objects of interest (Universe
  of Discourse), e.g. for ontology engineering

10
Without ontologies ...

• How do I know how to interpret
• Where do you come from ? (domain ambiguity)
• Geneva (the airport I started from) ?
• Lausanne or Switzerland (the place Im living in)
  ?
• France (the country I am a citizen of) ?
• Milano (the place I was born) ?
• Have a cup of coffee (context dependent)
• Would you consider paying 10000 US to buy a bad
  painting ? (term ambiguity)

11
Ontological Agreement

• Simple case common, shared ontology
• Needs services to define, store, retrieve,
  update, the ontology

12
Cooperative Systems

• Autonomous ontologies

Mediation Ontology
Ontology B
Ontology A
Mediator
A
B
information exchange
13
What's in an Ontology?

• Ontologies typically include
• Terms (names) for the important concepts in the
  domain
• Participant is a concept whose members are a kind
  of Person
• Author is a concept whose members are exactly
  those persons who write published papers
• StudentParticipant is a concept whose members are
  exactly those participants whose position is
  "student"
• formally, these sentences are expressed as axioms
  defining the new concepts- Participant is
  defined as a subconcept of Person - Author is
  defined as a restriction of Person based on the
  write role associated to persons-
  StudentParticipant is defined as a restriction of
  Person based on the position role associated to
  persons

14
Visualization (Protégé)

15
What's in an Ontology?

• Background knowledge (general rules)Constraints
  on the domain
• StudentParticipants pay a reduced registration
  fee
• StudentParticipants must have a supervisor
• No individual can be both a Reviewer and an
  Author for the same paper
• Instances/Individuals
• Stefano Participant
• Stefano Author

Terminological axioms
Assertional axioms
16
Evolution of Ontologies

• Terms
• Terms properties reasoning
• Terms properties reasoning space time

Taxonomic Ontologies
Descriptive Ontologies
ST Descriptive Ontologies
17
Taxonomic Ontologies

• sophisticated dictionary/thesaurus
• organized collection of terms
• some semantic links (synonymy, etc.)
• generalization/specialization hierarchy
• example Wordnet
• They provide a reference vocabulary

18
Wordnet

• 1 entity, physical thing
• 1 object, physical object
• 1 living thing, animate thing
• 1 organism, being
• 1 animal, animate being, beast, brute,
  creature, fauna
• 1 ....
• 1 mammal
• 1 placental mammal, eutherian
• 1 ungulate, hoofed animal
• 1 odd-toed ungulate, perissodactyl
• 1 equine, equid
• 1 horse, Equus, caballus
• 1 saddle horse,

Horse a solid hoofed herbivorous
quadruped domesticated since prehistoric times
19
Descriptive Ontologies

• concepts are worth a description
• beyond how to denote them (terms)
• which characteristic properties?
• which characteristic relationships?
• They provide
• description of information that may be available
  on the concept
• information to align existing data structures
• patterns to define new specialized ontologies
• -- same as conceptual modeling?

Lake

name (1n)
f
depth
( ) (11)
harbors (0n) beaches (0n) incomingRiver(0n) out
going river islands (0n) ........
20
Ontologies and DB

• Similarities, but important differences in
  their goals
• DB
• target data management, for a given organization
• prescribe how the world of interest is (Closed
  World Assumption)
• Ontologies
• target data description, for the largest
  community
• describe what is known about the real world (Open
  World Assumption)

21
CWA / OWA reasoning

• Parent instances of Person that have at least
  one child
• DL
• john Parent john is an instance of Person
  john has at least one role hasChild
  (unknown)
• DB john Parent rejected

DL
DB
Person
hasChild
Parent
22
CWA / OWA reasoning

• DL john has at most 2 children
• john has a child peter
• john has a child mary
• john has a child paul
• (paul should be the same as peter) or (paul
  should be the same as mary) or (mary should be
  the same as peter)
• Open world assumption no unique name
  assumption
• (but implemented reasoners have unique name
  assumption)
• DB john has at most 2 children
• john has a child peter
• john has a child mary
• john has a child paul rejected

23
Ontology Design

• In a web services perspective ...
• Ontology design is collaborative
• Ontology design is incremental
• needs reasoning services
• to check consistency of the specifications
• to accurately integrate new knowledge
• to infer all inferable knowledge
• Formal reasoning Logics

24
Consistency checking

• Satisfiability
• Parent ? (Person ? ? hasChild.Person)
• Woman ? (Person ? Female)
• Mother ? (Female ? Parent)
• GayMother ? (? Woman ? Mother)
• GayMother can never be satisfied (instantiated)
• A concept C is satisfiable iff there exists an
  interpretation I such that CI ? Ø (I is called
  a model of C)

25
Subsumption reasoning

• concept Person
• role citizenOf (Person, Country)
• role livesIn (Person, Country)
• Japanese ? (Person ? ? citizenOf.Country
  Japan)
• JapaneseDiaspora ? (Person ? (?
  citizenOf.Country Japan
  ? ? livesIn.Country ? Japan)
• Also at the instance levelHitori
  JapaneseDiaspora Hitori Japanese

26
Subsumption reasoning

• concept Person
• role hasAge (Person, Integer)
• Adult ? (Person ? ? hasAge 30)
• Senior ? (Person ? ? hasAge 60)
• NOT POSSIBLE

27
Subsumption reasoning

• concept Person
• Parent ? (Person ? ? hasChild.Person)
• Woman ? (Person ? Female)
• Mother? (Parent ? Female)
• subsumption can be reduced to satisfiability

28
Inference Case reasoning

• Does John have a female friend loving a male
  person?

29
Logics

• First Order Logic (FOL) is not decidable
• Description Logics are decidable subsets of FOL
• no free variables
• axioms knowledge representation and reasoning
• problem scalability of DL reasoners
• Horn-Logic is another decidable subsets of FOL
• only one negation
• rules deduction
• F-Logic a Horn-Logic supporting frames

30
Description Logics

• Decidable subsets of FOL (no free variables)
• Designed for knowledge representation and
  ontological reasoning
• Many variants (different compromises between
  expressive power, decidability, and complexity of
  reasoning)
• Very popular with the AI-Ontology community
• Focuses on axiomatic description of concepts and
  roles (T-box), but also allows description of
  instances (A-box)
• DAML OIL, OWL, Racer, Fact, Protégé, ...

31
DL basic constructs

• Concept
• Human, Animal
• Role (oriented binary relationship)
• cyclic roles may be symmetric, transitive
• a role may have an inverse
• Generalization hierarchies concepts and roles
• Domain of values

32
Concept constructors

• intersection, union, complement
• Man ? (Human ? Male)
• Female ? Human ? (? Male)
• existential and universal quantifiers
• ? hasChild.Manx ?(y) (hasChild(x,y) -
  Man(y)
• AllSonsFather ? (Man ? ? hasChild.Man )
• minimum and maximum cardinality
• 1hasChild
• Father ? (Man ? 1hasChild )

33
Visualization

34
DL examples

• Primitive concepts
• Person, Committee, Paper, Conference
• Defined concepts
• Participant ? Person ? ? registers.Conference
• Author ? Person ? ? writes.Paper
• RegisteredAuthor ? Author ? ? registers.Conference
  
• GoodConference ? Conference ? ?
  chairedBy.Tanaka
• Constraints
• a committee has at least 10 members
  Committee ? 10 hasMember.Person

35
Horn-Logic

• Decidable subsets of FOL only one negation
• Rules a formalism for deduction (? axioms)
• Powerful support recursive rules
• Mostly relational-based (e.g., Datalog)
• reviewer(P,C) pcMember(P,C)
• reviewer(P,C) delegates(Px,P),
  reviewer(Px,C)

36
F-Logic

• Rule language, designed for deduction
• Object-oriented expressiveness
• person name personName,
• firstNames personFirstname,
• address_at_(type) personAddress,
• isMemberOf committee,
• chairs committee
• Rule Pchairman PpersonchairsC

37
Assessment

• DLs
• open world assumption
• automatic consistency checking
• automatic placement of new concepts
• good for distributed asynchronous coordination
• counter-intuitive
• poor expressiveness
• poor readability
• poor query languages
• poor scalability

38
Assessment (2nd)

• Horn-Logic and F-Logic
• close world assumption (mostly)
• no need for consistency checking
• no automatic placement of new concepts (no
  subsumption reasoning)
• can serve as an implementation platform for DL
• (cf. DIP project)

39
Practical Achievements

• Description Logics
• OWL Ontology Web Language (successor to
  DAMLOIL, RDF, ...)
• RACER a reasoning system for OWL (implements
  the SHIQ Logic)
• Interface Tools Protégé, OntoEdit, ..........
• Horn-Logics deductive DBMS
• F-Logic Florid, Ontobroker, Flora
• Outsiders
• KAON an ontology and semantic web framework,
  allowing the design and management of ontologies,
  but with limited reasoning capabilities
• DOGMA an ontology engineering framework based on
  the ORM (Object-Role-Modeling) conceptual model,
  with no reasoning capabilities

40
Our proposal

• Modeling-oriented (rather than reasoning-oriented)
  approach to ontologies
• Use Conceptual Modeling for expressiveness and
  understandability
• MADS spatio-temporal conceptual data model
• Enhance the capabilities of conceptual models to
  support reasoning
• symmetric and transitive cyclic relationships
• derived objects
• membership predicates
• ......
• Use Logic-based approaches for reasoning
• Hybrid System
• DBMS scalability
• DL-reasoners inferencing

41
MADS

• MADS a spatio-temporal conceptual model
  (complex objects, n-ary relationships with
  attributes, generalization hierarchies,
  multi-instantiation, spatio/temporal and
  contextual features)
• spatial objects geometry attribute
• spatial attributes
• spatial data types Point, Line, OrientedLine,
  Area, PointSet, LineSet, OrientedLineSet,
  AreaSet, Geo
• spatial relationships
• space-varying attributes (functions)
• The MADS framework includes a visual schema
  editor and a visual query editor (MurMur EEC/IST
  Project)

42
A MADS Spatial Schema

43
Object type example

44
Rationale for hybrid DBDL systems

• Descriptive ontologies require
• Rich models to enable building representations as
  close as possible of human perception
• Support for the precise definition of concepts in
  relation to other concepts
• Storage and transactions management mechanisms
  (security, concurrency, reliability) to
  realistically manage large sets of instances
• Both open world and closed world reasoning
• Query languages for schema exploration, reasoning
  on the schema, and querying of instances

45
Problem Dissimilarities

• Inheritance links
• DB currently do not provide a way to express what
  is the specialization criterion that defines a
  sub-class no possibility of automatically
  positioning an instance in a sub-class
• Multi-instantiation
• Default rules are different
• DL by default any two concepts may share
  instances
• MADS by default two types do not share instances
  
• Spatio-temporal information
• DL very limited support
• MADS good support discrete and continuous
  views, ST relationships

46
Defined Concepts

• Goal support the precise definition of concepts
  in relation to other concepts
• DL "defined concepts" by a logic formula
• defined concepts are managed in the same way as
  primitive concepts
• users may insert instances in defined concepts
• a StudentParticipant a.position
  "student"
• DB views and derived objects are the only way to
  define new subsets of instances
• Views are not part of the schema
• Views are not instantiable by users
• MADS derived spatial relationships and derived
  attributes

47
Instances

• Instance management principles
• Open / closed world assumption
• Oid / names (DL no unique name assumption)
• Instance management efficiency
• DBMS have been designed to provide storage and
  transactions management mechanisms (security,
  concurrency, reliability)
• DL does not have such facilities
• Constraints DL formulae can express constraints
  at both levels
• the schema (Tbox) level
• the instances level
• john has at most 2 children
• But formulae are used for inference and not for
  constraining data (OWA)

48
Conclusion

• Knowledge sharing is the issue
• Ontologies enable Semantic Web Services
• Ontologies need DB technology
• It is a long way to go, but
• it is THE way to go

49
Thanks for your attention