Part II: Representation

1
Part II Representation

• Thesauri, Topic Maps
• Frames, RDF/RDF-Schema
• DAMLOIL, Reasoning
• Services vs Data Structures

2
Ontology Representation

• Ontologies are the cornerstone of encoding
  understanding, BUT to be shared they they need a
  standard representation and exchange language

3
Language desirable properties
Machine communication

• Support for
• Multiple classification
• Can say simple things simply but as complex as
  necessary
• Expressive enough to capture many ontologies
• Evolution Merging
• Multiple authoring
• Web enabled coding to RDF/XML

Model or proof theory Tractability of reasoning
Strong conventions of use Human readable names

• Natural primitives

Human communication
4
Ontology Representation Languages

• Machines need communication with formal content
  to restrict meaning
• What makes a language formal?
• model theory (1st order predicate logic)
• proof theory (Gentzen calculus)
• conventions (e.g. Java)

5
Ontology its encoding rep.

• Ontology
• Domain specific conceptualisation expressed
  within some representational model
• Representational languages
• Data structuring mechanism in which ontology is
  expressed
• E.g. relational model, o-o model, frames, logics
• Ontology can be delivered as a (static) data
  structure for embedding in an application or a
  (dynamic) service.

API
6
XML for KR

• Definition of self-describing data in worldwide
  standardized, non-proprietary format.
• Structured data and knowledge exchange for
  enterprises in various industries.
• Integration of information from different sources
  to uniform documents.
• Exchange of knowledge bases between different AI
  languages, knowledge bases and databases,
  application systems, etc.
• But.

7
XML is not enough
The Creator of the Resource http//www.w3.org/Ho
me/Lassila is Ora Lassila

• XML defines grammars to verify and structure
  documents
• The grammar enforces constraints on tags
• Different grammars define the same content
• XML lacks a semantic model it only has a
  surface model which is a tree.

8
XML is not enough

• Meaning of XML documents is intuitively clear
• semantic markup tags are domain terms
• But computers do not have intuition
• Tag names per se do not provide semantics
• The semantics are encoded outside the XML
  specification
• XML makes no commitment on
• Domain specific ontological vocabulary
• Ontological modelling primitives
• ? requires pre-arranged agreement on ? ?
• Feasible for closed collaboration
• agents in a small stable community
• pages on a small stable intranet

9
Representation Paradigms (incomplete)
TopicMaps
Thesauri
Taxonomies
Semantic Nets
Ontologies
Extended ER-Model/UML Db schema
Logics Predicate Logics F-Logic Conceptual
Graphs Description Logics
RDF(S)
Frames
Expressivity
10
Representation Paradigms (incomplete)
TopicMaps
Thesauri
Taxonomies
Semantic Nets
Ontologies
Extended ER-Model/UML
Logics Predicate Logics F-Logic Conceptual
Graphs Description Logics
RDF(S)
Frames
11
Thesauri
similarTo
Fruit
Vegetable
Example
Apfelsine (german)
NarrowerTerm
Orange
synonymWith

• Graph with labels edges (similar, nt, bt,
  synonym)
• Fixed set of edge labels (aka relations)
• no instances

• Well known in library science
• cf. terminologies / classifications (Dewey)

12
Vocabularies Gene Ontology

• Hand crafted with simple tree-like structures.
• Position of each concept and its relationships
  wholly determined by a person.
• Flexible but
• Maintenance and consistency preservation
  difficult and arduous.
• Poor semantics.
• Single hierarchies are limiting
• cross products.

13
Topic Maps (I)

• Standardized ISO/IEC 132502000 (Jan 2000)
• enabling standard to describe knowledge
  structures,electronic indices, classification
  schemes, ...
• To enable information resources to be classified
  and navigated in a consistent manner by
  representing knowledge structures for indexes
• build valuable information networks above any
  kind of resources / data objects
• enable the structuring of unstructured
  information
• To make subjects addressable.
• The GPS of the Web

http//www.topicmaps.org
14
Back-of-the-Book Index British Virgin Islands
Gorda Sound see North Sound Little Dix Bay
.................... 89 North Sound
....................... 90 Road Harbour see also
Road Town ... 73 Road Town ......................
69,71Spanish Town ................... 81,82 Torto
la ........................... 67Virgin Gorda
...................... 77
15
Topic Maps (II)

• The electronic equivalent of table of contents,
  glossaries, thesauri, cross references
• Subjects become objects ("topics").
• Relationships between subjects are asserted.
• The concepts or topics that underlie a set of
  information objects exposed to those people or
  applications processing the information.

16
High-level Topic Maps concepts
Knowledge, Occurrences, Associations, Names
Association
Topic
-Roles -Members -Template
-Subject -Name(s)
Graph made of nodes and arcs, based on Semantic
Nets Typed to form groups (types are defined in
the standard as topics) Topics and Occurrences
play roles in relationships Based on Platos
notions.
Occurrences
-Type
Scope
17
In-/Semi-formal approaches Topic Maps, Thesauri

• Advantages
• Capture a lot of modelling experiences.
• Intuitive.
• Interesting primitives that are not available in
  other approaches (TM).
• Topic Maps Web enabled
• XML Topic Maps (XTM) are ready to use.

• Disadvantages
• No characterization independent from particular
  implementation.
• May be misinterpreted (TM) / few primitives
  (Thesauri).
• No formal interpretation.
• No formal rigour.
• Hard to build and maintain large and coherent
  schemes.
• Pre-enumerate concepts.

18
Representation Paradigms (incomplete)
TopicMaps
Thesauri
Taxonomies
Semantic Nets
Ontologies
Extended ER-Model/UML
Logics Predicate Logics F-Logic Conceptual
Graphs Description Logics
RDF(S)
Frames
19
Frames, SDM, OO models

• Frames
• Rich set of language constructs frames, slots,
  facets, defaults.
• Impose restrictive constraints on how they are
  combined or used to define a class.
• All frames asserted into taxonomy by hand.
• All concepts are primitive.
• Octet/GKB, Protégé, OCML, Ontolingua
• OKBC Open Knowledge Base Connectivity.
• OKBC-Lite.
• OO / Semantic Data Models (EER, UML)
• Taxonomy/inheritance semantics?
• Intuitive, lots of tools, widely used.

20
Frame Data Model

• Frames
• Classes Genes, Reactions
• Instances
• Relationships
• Slots Chromosome, map-position, citations,
  reactants, products, Keq
• Facets Chromosome is single-valued, instance of
  class Chromosomes Citations is multiple valued,
  set of strings.

21
Protégé 2000
http//www.smi.stanford.edu/projects/protege/
22
RDF Web based data model

• Semantic Web beyond machine readable to machine
  understandable.
• Resource Description Framework is the W3C
  language for describing metadata on the Web.
  http//www.w3.org/RDF
• RDF consists of two parts
• RDF Model (a set of triples)
• RDF Syntax (different XML serialization syntaxes)
• RDF a small set of modelling primitives syntax
• RDF does not commit to a domain vocabulary
• RDF Schema for definition of Vocabularies (simple
  Ontologies) for RDF (and in RDF)

23
A simple RDF example

• Resources
• A thing you can reference (URI)
• RDF definitions are themselves Resources.
• Properties
• slots, defines relationship to other resources or
  atomic values
• Similar to Frames.
• Statements
• Resource has Property with Value
• Values can be resources or atomic XML Schema data
  types.
• Directed graph

http//www.w3.org/Home/Lassila
sCreator
Ora Lassila
Triples Resource (subject)
http//www.w3.org/Home/Lassila Property (predicate
) http//www.schema.org/Creator Value
(object) "Ora Lassila
24
Collection Containers

• Multiple occurrences of the same PropertyType
  doesnt establish a relation between the values
• The Millers own a boat, a bike, and a TV set
• RDF defines three special Resources
• Bag
• Sequence
• Alternative

RdfBag
rdftype
/Students/Amy
students
rdf_1
rdf_2
/Students/Tim
bagid1
rdf_3
/Students/John
rdf_4
/Students/Mary
rdf_5
The students incourse 6.001 are Amy, Tim,John,
Mary, and Sue
/Students/Sue
25
Statements about statements

• Transform them into Resources.
• Ralph Swick believes that
• the creator of the resource http//www.w3.org/Home
  /Lassila is Ora Lassila

26
RDF Schema (RDFS)

• RDF just defines the data model.
• Need for definition of vocabularies for the data
  model - an Ontology Language!
• RDF-Schemas describe rules for using RDF
  properties
• Define a domain vocabulary for RDF
• Organise this vocabulary in a typed hierarchy
• RDF Schemas are Web resources (and have URIs) and
  can be described using RDF.
• Are not to be confused with XML Schemas.
• RDFS is the framework for a vocabulary.

27
RDF Schema Model

• Property-centric Each property specifies what
  classes of subjects and objects it relates. New
  properties can be added to a class without
  modifying the class
• resource, class, subClassOf, type
• property, subPropertyOf
• domain, range, constraintResource,
  constraintProperty
• Definitions can include constraints which express
  validation conditions
• domain constraints link properties with classes
• range constraints limit property values
• BUT expressive inadequacy and poorly defined
  semantics

28
RDF Schema Model
Resource
Range
subClassOf
Person
type
Property
Class
type
type
MotorVehicle
registeredTo
type
type
subClassOf
subClassOf
Domain
Truck
ownedBy
29
Frame/OO model summary

• Advantages
• Intuitive and popular modelling style.
• Many tools and examples.
• OKBC standard for semantics.
• Some reasoning.

• Disadvantages
• Extending/evolving problematic
• Hand crafting taxonomies and asserted properties.
• Static classifications.
• Pre-enumerate concepts.
• Little reasoning support
• Difficult to build large coherent and complete
  ontologies (e.g. multiple classifications)

30
RDF Resources

• RDF Repositories
• RDFDB, RDFSuite, Sesame
• RDF Query Languages
• RQL, RDQL, SQUISH
• Annotation systems to create RDF
• Annotea, CREAM
• COHSE (see later)
• RDF Java API
• Jena http//www.hpl.hp.com/semweb/jena-top.html

http//www.ilrt.bristol.ac.uk/discovery/rdf/resour
ces
31
Representation Paradigms (incomplete)
TopicMaps
Thesauri
Taxonomies
Semantic Nets
Ontologies
Extended ER-Model/UML
Logics Predicate Logics F-Logic Conceptual
Graphs Description Logics
RDF(S)
Frames
32
RDF(S) Extensibility
Define an Ontology of your Language with RDF
Schema (like RDF-Schema itself) Describe Instance
Data using your new Vocabulary Advantage all
Languages use the same Data Model (simplifies
Interoperability)
Definition uses the Data model of RDF
RDF Schema
Defined in terms of
Is extension of
33
Stack of languages
34
The Ontology Language Stack
OWL
DAML-S
DAML-R
DAMLOIL
OIL
DAML-Ont
DC
PICS
RDF(S)
RDF
Topic Maps
SMIL
XOL
HTML
XML Name Space XML Schema
Unicode
URI
35
Web Language Stack summary

• XML
• interchange syntax, no semantics
• RDF
• Data model, some semantics inference (recent!)
• RDF Schema
• concept modelling, more semantics inference
• DAMLOIL / OWL
• more expressive ontology language
• quite expressive expensive inference
• Requirements for a Web Ontology Language W3C
• http//www.w3.org/TR/webont-req/

36
History DAMLOIL

• OIL developed by group of (largely) European
  researchers.
• DAML- ONT developed by group of (largely) US
  researchers (in DARPA DAML programme).
• Efforts merged to produce DAML OIL.
• Development was overseen by joint EU/ US
  committee.
• Now submitted to W3C as basis for standardisation
  WebOnt working group developing language
  standard.
• Likely to be called OWL.
• http//www.w3.org/2001/sw/WebOnt/

37
DAMLOIL / OWL

• DAML OIL designed to describe structure of
  domain (schema)
• Object oriented classes (concepts) and
  properties (roles)
• DAMLOIL ontology consists of set of axioms
  asserting characteristics of classes and
  properties
• E.g. Person is kind of Animal whose parents are
  Persons
• RDF used for class/property membership assertions
  (data)
• E.g. John is an instance of Person h John
  Mary i is an instance of parent

38
DAMLOIL / OWL

• DAML OIL supports the full range of XML Schema
  data types
• Primitive (e. g., decimal) and derived (e. g.,
  integer sub- range)
• DAML OIL classes can be names (URIs) or
  expressions
• Various constructors provided for building class
  expressions
• Expressive power determined by
• Kinds of constructor provided
• Kinds of axiom allowed

39
Description Logics

• DAML OIL equivalent to the expressive
  Description Logic (an extension of) SHIQ DL
• The descendants of frame systems and object
  hierarchies via KL-ONE.
• Core distinction between class definitions (T-Box
  ? Schema) and instance definitions (A-Box ?
  Database tuples)
• Many years of DL research

• Well defined semantics
• Formal properties well understood (complexity,
  decidability)

• Known reasoning algorithms
• Implemented systems (highly optimised)

40
Whats in a Logic based ontology?
Rector

• Primitive concepts - in a hierarchy
• Described but not defined
• Properties - relations between concepts, also in
  a hierarchy
• Constructors on concepts and properties
• some, only, at least, at most, and, or,
  not.
• Defined concepts
• Made from primitive concepts, constructors and
  descriptors
• Enzyme ? protein and catalyses reaction.
• Reason that enzyme is a kind of protein.

• is-kind-of implies
• Dog is a kind of wolf mean All dogs are
  wolves
• Axioms
• disjointness, further description of defined
  concepts
• A Reasoner
• to organise it for you. Consistency Taxonomy
  for defined concepts established though logical
  reasoning.

Model built up incrementally and descriptively
based on concepts properties.
41
Logic Based Ontologies
Rector
Thing
red partOf Heart
red partOf Heart
(feature pathological)
42
(No Transcript)
43
(No Transcript)
44
Reasoning support

• Consistency check if knowledge is meaningful
• Subsumption structure knowledge, compute
  taxonomy
• Equivalence check if two classes denote same
  set of instances
• Instantiation check if individual i instance of
  class C
• Retrieval retrieve set of individuals that
  instantiate C
• Problems all recucible to consistency
  (satisfiability)

45
Reasoning demo using OilEd

• http//oiled.man.ac.uk

46
(No Transcript)
47
(No Transcript)
48
Semantics matters

• A hacker who studied ontology
• Was famed for his sense of frivolity
• When his program inferred
• That Clyde ISA Bird
• He blamed not his code but zoology
• Clyde ISA Elephant
• AI limericks by Henry Kautz
• http//www.cs.washington.edu/homes/kautz/misc/lime
  ricks.html

49
Why Reasoning Services I ?

• Ontology design
• Check class consistency and (unexpected) implied
  relationships
• Particularly important with large
  ontologies/multiple authors
• Ontology integration
• Assert inter-ontology relationships
• Reasoner computes integrated class
  hierarchy/consistency

• Ontology deployment
• Determine if set of facts are consistent w. r. t.
  ontology
• Determine if individuals are instances of
  ontology classes
• Query Inclusion
• Service description matchmaking
• Classification-based querying.

50
Gain of mapping?

• Any RDF agent can process DAMLOIL instances
• Any RDF-S agent can process DAMLOIL ontologies
• Any DAMLOIL-aware agent can exploit semantics
  reasoning(and materialize the DAMLOIL
  derivations for use by DAMLOIL-ignorant RDF
  agents)

51
Evaluation of DAMLOIL /OWL

• Advantages
• Decidable (if choosen carefully like DAMLOIL)
• Subsumption reasoning
• Consistency checking
• Dynamically post-coordinate rather than have to
  pre-enumerate.
• Support for evolution, merging, large scale
  building
• Can publish the ontologies as static lattices.
• W3C standard (so tools etc)

• Disadvantages
• Different modeling style if want to take
  advantage of reasoning.
• Limited support for A-Box reasoning (on
  instances) in tractable DL versions

As simple as required but as complex as
necessary
52
Languages Summary

• Thesauri Topic Maps
• Hand crafted, flexible but difficult to evolve,
  maintain and keep consistent, with poor
  semantics.
• Object-based KR e.g. frames
• Extensively used, good structuring, intuitive.
• Semantics defined by OKBC standard.
• Logic-based Description Logics
• Very expressive, model is a set of theories, well
  defined semantics, reasoning.
• Automatic derived classification taxonomies.
• Concepts are defined and primitive.
• Expressivity vs. computational complexity balance.

53
Common language errors

• AI peoples errors
• it is good if it is formal
• it is good if someone with a logic background may
  easily use it
• it is good if the language allows everything

• Engineers errors
• it works in my application, thus it is good
• who needs formality anyway?
• it did not work when I looked at it 10 years ago

54
Further Reading

• ltsupplied separatelygt