Overview
• Definition of ontology
Ontologies I
Fredrik Arvidsson
LIBLAB/HCS/IDA
Annika Flycht-Eriksson
NLPLAB/HCS/IDA
• Perspectives on ontologies
• Ontologies for the Semantic Web
• Design and development of ontologies –
methods and tools
Ontologies I
What is an ontology?
What is an ontology?
In practice…
In theory…
• Simple concept hierarchies
• Semantic nets
Fredrik Arvidsson & Annika Flycht-Eriksson
”An ontology is a formal, explicit
specification of a shared conceptualisation.”
• Frame systems
(Gruber, 1993)
• Logical models
Ontologies I
Fredrik Arvidsson & Annika Flycht-Eriksson
Ontologies I
What is an ontology?
Types of ontologies
In English...
From general to specific
An ontology provide a shared vocabulary,
which can be used to model a domain, that
is, the type of objects and/or concepts that
exist, and their properties and relations
Fredrik Arvidsson & Annika Flycht-Eriksson
• Generic
• Core
• Domain
• Task
• Application
Ontologies I
Fredrik Arvidsson & Annika Flycht-Eriksson
Ontologies I
Fredrik Arvidsson & Annika Flycht-Eriksson
1
The anatomy of ontologies
Perspectives
• Terms
• Philosophy
• Definitions of terms
• Library and Information Science
• Axioms
• Artificial Intelligence
• (Inference mechanisms)
• Natural Language Processing
• The Semantic Web
Ontologies I
Fredrik Arvidsson & Annika Flycht-Eriksson
Ontologies I
Fredrik Arvidsson & Annika Flycht-Eriksson
Philosophy
Library & Information Science (LIS)
• Objectives
• Objectives
• Describe world, classify and categorise
• Examples
• Organise the bibliographic universe
• Model universal & domain knowledge
• Usage
• Aristotle, …, Sowa
• Provide accesspoints to bibliographic entities
• Collocation device
• Examples
• Dublin Core, MARC
• LCC, UDC, SAB
Ontologies I
Fredrik Arvidsson & Annika Flycht-Eriksson
Ontologies I
Fredrik Arvidsson & Annika Flycht-Eriksson
Artificial Intelligence
Natural Language Processing
• Objectives
• Objectives
• Model common sense and domain knowledge
• Usage
• Knowledge representation and reasoning
• Examples
• Machine Translation, Information Extraction, Q/A
• Examples
• CYC, ...
Ontologies I
• Model lexical and domain knowledge
• Usage
• Wordnet, Generalised Upper Model, Dahlgren
Fredrik Arvidsson & Annika Flycht-Eriksson
Ontologies I
Fredrik Arvidsson & Annika Flycht-Eriksson
2
The Semantic Web
The Ontology Web Language - OWL
• Objectives
Goal
• Provide semantics for web resources
• Usage
• Describe resources
…and their contents
• Go beyond the basic semantics in RDFS
Current status:
• Examples
•
•
•
•
• DC, DAML-library, …
Ontologies I
• Formally describe the semantics of classes
and properties used in web documents.
Fredrik Arvidsson & Annika Flycht-Eriksson
use cases and their requirements on ontologies
eight design goals
19 requirements
…and some objectives
Ontologies I
Fredrik Arvidsson & Annika Flycht-Eriksson
Design goals
Requirements
• Shared ontologies
• Ontologies as distinct objects
• Ontology evolution
• Unambiguous term referencing with URIs
• Ontology interoperability
• Explicit ontology extension
• Inconsistency detection
• Commitment to ontologies
• Balance of expressivity and scalability
• Ontology metadata
• Ease of use
• Versioning information
• XML syntax
• Class definition primitives
• Internationalisation
• Property definition primitives
Ontologies I
Fredrik Arvidsson & Annika Flycht-Eriksson
Ontologies I
Fredrik Arvidsson & Annika Flycht-Eriksson
Requirements, cont.
Requirements, cont.
• Datatypes
• User-displayable labels
• Class and property equivalence
• Supporting a character model
• Individual equivalence
• Supporting a uniqueness of Unicode strings
• Local unique names assumption
• Attaching information to statements
• Classes as instances
• Complex data types
• Cardinality constraints
Ontologies I
Fredrik Arvidsson & Annika Flycht-Eriksson
Ontologies I
Fredrik Arvidsson & Annika Flycht-Eriksson
3
Design and development
Design and development approaches
• Design guidelines and principles
• Inspirational
• Guarino, Gruber,...
• Inductive
• Methods
• Deductive
• Methonology, TOVE,...
• Synthetic
• Tools
• Protégé, KAON,...
• Collaborative
• Libraries
• Ontolingua server, DAML-library,...
Ontologies I
Fredrik Arvidsson & Annika Flycht-Eriksson
Ontologies I
Fredrik Arvidsson & Annika Flycht-Eriksson
General development methodology
Ontology learning
• Specify purpose and scope
• ML and NLP techniques to facilitate capture
and organisation
• Capture, define, organise
• Extend/Refine existing ontology
• Formalise, implement
• Integrate existing resources
• Free or semi-structured text as source
• Evaluate
Ontologies I
Fredrik Arvidsson & Annika Flycht-Eriksson
Ontologies I
Text-To-Onto (OntoEdit)
Dublin Core
• Import and Reuse
Goals
• Extract
•
•
•
•
Lexical entry and concept extraction
Hierarchical concept clustering
Dictionary parsing
Association rules
• Provides a semantic vocabulary for describing the
“core” information properties of resources (electronic
and “real” physical objects)
• Proviede enough information to enable intelligent
resource discovery systems
History
• A collaborative effort started in 1995
• Initiated by people from computer science,
librarianship, online information services, abstracting
and indexing, imaging and geospatial data, museum
and archive control.
• Prune
• Refine
Ontologies I
Fredrik Arvidsson & Annika Flycht-Eriksson
Fredrik Arvidsson & Annika Flycht-Eriksson
Ontologies I
Fredrik Arvidsson & Annika Flycht-Eriksson
4
Dublin Core - 15 Elements
DC - Classes of qualifiers
15 Elements related to a resource:
Two broad classes of qualifiers:
• Content
• Element Refinement
• Title, Subject, Description, Type, Source,
Relation and Coverage
• ”…make the meaning of an element narrower or more
specific.”
• HTML example from ”www.kb.se”:
<META NAME="DC.Date.Modified" CONTENT="1999-06-02">
• Intellectual property
• Creator, Publisher, Contributor, Rights
• Instantiation
• Date, Language, Format, Identifier
• Encoding Schema
• ”…identify schemas that aid in the interpretation of an
element value.”
• HTML example form “www.kb.se”:
<META NAME="DC.Subject" SCHEME="SAB”
CONTENT="Nationalbibliotek">
Ontologies I
Fredrik Arvidsson & Annika Flycht-Eriksson
Ontologies I
Fredrik Arvidsson & Annika Flycht-Eriksson
Questions and Research directions
Design & development:
• Consensus / Collaborative v.s. Individual
• Level of granularity
Control & use:
• Centralised v.s. distributed
• Interoperability
The OWL objectives
• Layering, Commitment to portions of ontologies
• Default values, CWA, Procedural attachment
• ...
Ontologies I
Fredrik Arvidsson & Annika Flycht-Eriksson
5