Welcome to Ontology Engineering Guus Schreiber
Welcome to Ontology Engineering
Guus Schreiber
Agenda • Course introduc:on: what is an ontology? • Administra:on
• RDF/RDFS
Lecture 1
Literature
• James Odell, Ontology White Paper, CSC Catalyst, 2011, V2011-‐07-‐15,
hNp://www.jamesodell.com/Ontology_White_Paper_2011-‐07-‐15.pdf.
• For this lecture Sec.s 1-‐4 are relevant
• Acknowledgement: some figures in this lecture come from the paper above.
What is an Ontology?
• In philosophy: theory of what exists in the world • In IT: consensual & formal descrip:on of shared concepts in a domain • Aid to human communica:on and shared understanding, by specifying meaning
• Machine-‐processable (e.g., agents use ontologies in communica:on)
• Key technology in seman:c informa:on processing • Applica:ons: knowledge management, e-‐business, seman:c world-‐wide web.
What is an Ontology? II
“explicit specifica-on of a shared conceptualiza-on that holds in a par-cular
context” (several authors)
Knowledge sharing and reuse
• Knowledge engineering is costly and :me-‐consuming
• Distributed systems
• Increasing need for defini:on of a common frame of reference – Internet search, document indexing, ….
Need for data integra:on?
Seman:c Web
• Data integra:on • AAA slogan • Non-‐Unique Naming Assump:on
• Open vs. closed World
The Web: resources and links
URL URL
Web link
The Seman:c Web: typed resources and links
URL URL
Web link
ULAN
Henri Ma:sse
Dublin Core
creator
Pain:ng “Woman with hat SFMOMA
Seman:c Web
WordNet
14
Domain standards and vocabularies as ontologies
• Contain ontological informa:on • Ontology needs to be “extracted”
– Not explicit • Lists of domain terms are some:mes also called “ontologies” – Implies a weaker no:on of ontology – Scope typically much broader than a specific applica:on domain
– Contain some meta informa:on: hyponyms, synonyms, text
• Structured knowledge is available (on the web) – use it!
Ontology spectrum
Source: http://www.jamesodell.com/Ontology_White_Paper_2011-07-15.pdf.
16
Document fragment ontologies
17
Instruc:onal document fragment ontologies
Context and Domain
Principle 1: “The representa:on of real-‐world objects always depends
on the context in which the object is used. This context can be seen as a “viewpoint” taken on the object. It is usually impossible to enumerate in advance all the possible useful viewpoints on (a class of ) objects.”
Principle 2: “Reuse of some piece of informa:on requires an explicit
descrip:on of the viewpoints that are inherently present in the informa:on. Otherwise, there is no way of knowing whether, and why this piece of informa:on is applicable in a new applica:on seing.”
19
Mul:ple views on a domain
• typical viewpoints captured in ontologies: • func:on • behavior, • causality • shape, geometry • structure: part-‐of (mereology), aggrega:on • connectedness (topology)
• viewpoints can have different abstrac:on (generaliza:on) levels
• viewpoints can overlap • applica:ons require combina:ons of viewpoints
20
Mul:ple views on a domain
21
Context specifica:on through ontology types
• Domain-‐specific ontologies – Medicine: UMLS, SNOMED, Galen – Art history: AAT, ULAN – STEP applica:on protocols
• Task-‐specific ontologies – Classifica:on – E-‐commerce
• Generic ontologies – Top-‐level categories – Units and dimensions
22
Top-‐level categories: many different proposals
Chandrasekaran et al. (1999)
The famous is-‐a rela:onship
Source: http://www.jamesodell.com/Ontology_White_Paper_2011-07-15.pdf.
Classes as instances
24 Source: http://www.jamesodell.com/Ontology_White_Paper_2011-07-15.pdf.
What is an Ontology?
“explicit specifica-on of a shared conceptualiza-on that holds in a par-cular
context” (several authors)
Concepts
• Help us organize the world around us • Act as recogni:on device • Test for reality • We use many different types of concepts
Concept types
Source: http://www.jamesodell.com/Ontology_White_Paper_2011-07-15.pdf.
The concept triad
Source: http://www.jamesodell.com/Ontology_White_Paper_2011-07-15.pdf.
Concept specifica:on
• Symbol – Name used for the concept – Can be different names, different languages – E.g., “bike”, fiets”
• Intension (defini:on) – Intended meaning of the concept (seman:cs) – E.g. a bike has at least one wheel and a human-‐powered movement mechanism
• Extension – Set of examples of the concept – E.g. “my bike”, “your bike”
Incomplete concept specifica:ons
• Are common • Think of an example:
– Concept with no instances – Concept with no symbol
• Primi:ve vs. defined concepts
Domain = area of interest
• Can be any size – e.g., medicine
• Concepts may have different symbols in different domains
• The same symbol may be used for different concepts in different domains (some:mes also in the same domain)
Ontology Specifica:on
• Aggrega:on • Rela:on-‐aNribute dis:nc:on • Trea:ng rela:ons as classes • Sloppy class/instance
dis:nc:on – Class-‐level aNributes/rela:ons
– Meta classes • Constraints • Data types • Modularity
– Import/export of an ontology
• Ontology mapping
• Class (concept) • Subclass with inheritance • Rela:on (slot)
33
Ontology Languages
– UML – RDF Schema, OWL – …..
• Common basis – Class (concept) – Subclass with inheritance – Rela:on (slot)
Ontology Tools
Best known tool • Protégé (Stanford) • We will use this tool
Decision points: – Expressivity – Graphical representa:on – DB backend – Modulariza:on support – Versioning
Administra:on
• Course website: hNp://seman:cweb.cs.vu.nl/OE2012/
• Use blog posts for content ques:ons • Use oe-‐[email protected] for admin ques:ons
Engineering needs prac:ce!
Lots of exercises throughout the course: • Two mee:ngs per week
• Lectures on Monday
• Work sessions on Thursday
• You are encouraged to do assignments together with colleagues
• Individual porsolio
RDF(S) Recap
• Which RDF/RDF-‐Schema constructs do you remember?
URIs, URLs
• URI: global iden:fier for a web resource • hNp://www.w3.org/2006/03/wn/wn20/instances/synset-‐anniversary-‐noun-‐1
• URL: dereferencable URI, used to locate a file on the web.
• hNp://www.w3.org/2006/03/wn/wn20/instances/synset-‐anniversary-‐noun-‐1
• URI abbrevia:ons: – Qnames
• Namespace:iden:fier • Wordnet:synset-‐anniversary-‐noun-‐1
Triples
ulan:Shakespeare ulan:parentOf ulan:Susanna.
kb:Hamlet kb:author kb:Shakspeare.
ex:VrijeUniversiteit ex:locatedIn tgn:Amsterdam.
ex:WillemHage ex:teaches ex:OntologyEngineering.
ex:OntologyEngineering rdf:type ex:Course.
Syntax
• N3 Turtle – hNp://www.w3.org/TeamSubmission/turtle/
• RDFXML – hNp://www.w3.org/TR/rdf-‐syntax-‐grammar/
Blank nodes
How would you model “Sonnet78 was inspired by a woman who lives in England”?
Lit:Sonnet78 lit:hasInspiration [ rdf:type bio:Woman; bio:livedIn geo:England ] .
subClassOf
IF A rdfs:subClassOf B r rdf:type A
THEN r rdf:type B
subPropertyOf
IF P rdfs:subPropertyOf R a P b
THEN a R b
Domain and Range
IF P rdfs:domain D x P y
THEN x rdf:type D
IF P rdfs:range R x P y
THEN y rdf:type R
More RDF(S)
• rdfs:label • rdfs:comment
• rdfs:seeAlso
RDF-‐Schema
• Provides a way to talk about the vocabulary – Define classes, proper:es
bb:author rdf:type rdfs:Property • Enables inferencing
– Inferring new triples from asserted triples.
• subClassOf, subPropertyOf, domain, range.
47
Guidelines for ontological engineering (1)
• Do not develop from scratch • Use exis:ng data models and domain standards as star:ng point
• Start with construc:ng an ontology of common concepts
• If many data models, start with two typical ones • Make the purpose and context of the ontology explicit – E.g. data exchange between ship designers and assessors
– Opera:onally purpose/context with use cases • Use mul:ple hierarchies to express different viewpoints on classes
• Consider trea:ng central rela:onships as classes
48
Guidelines for ontological engineering (2)
• Do not confuse terms and concepts • Small ontologies are fine, as long as they meet their goal • Don’t be overly ambi:ous: complete unified models are
difficult • Ontologies represent sta:c aspects of a domain
– Do not include work flow • Use a standard representa:on format, preferably with a
possibility for graphical representa:on • Decide about the abstrac:on level of the ontology early
on in the process. – E.g., ontology only as meta model