Ontology modelling and the semantic web - Presentation Transcript
Ontology modelling and the semantic web Asgeir Rekkavik Deichmanske bibliotek
Librarian / consultant at Deichmanske bibliotek (Oslo Public Library)
...since 2004.
Involved in ontology related projects since 2006
Los topic structure for indexing public services
Sublima tool for subject driven portals
Mediazone multimediastations
What does semantic mean?
Semantics : The branch of linguistics concerned with meaning.
( Shorter Oxford English dictionary )
Semantics is the study of meaning, usually in language.
(Wikipedia 2009-10-19)
I love you
I ♥ U
Different syntax, same semantics
What does ontology mean?
Ontology : The science or study of being.
( Shorter Oxford English dictionary)
In computer science and information science, an ontology is a formal representation of a set of concepts within a domain and the relationships between those concepts.
(Wikipedia 2009-10-19)
What does ontology mean?
The world can be described in many different ways: e.g. language, art etc.
An ontology describes the world in a way that is formal, structured and unambiguous.
Why? Because we want to describe it to computers.
Taxonomies
Hierarchical classification
Characteristics
Generic relations (’is-a’ relations)
Directed graph
Nodes represent categories
Arrows represent broader/narrower relations
Especially known from biology. Developed by Carl von Linné.
Taxonomies
Taxonomies
Other transitive relations can exist between concepts, e.g. ’part-of’ relations
Different types of relations
Generic (”is-a”, e.g. Cat - Animal)
Partitive (’part-of’, e.g. Oslo - Norway)
Instance (e.g. Sokrates - Philospher)
Associative (several)
In a thesaurus, the generic, partitive and instance relations are all expressed as broader / narrower. No distinction between different associative relations.
Thesaurus hierarchy
Generic class hierarchy
Generic class hierarchy
Prot égé
Free, open source ontology editor
Developed by Stanford University and the University of Manchester
Latest version Prot égé 4.0
Available from:
http://protege.stanford.edu
Currently developing: WebProt égé (Alpha)
Classes
Represent categories, sets of individual instances
Are related to eachother through parent-child relationships (superclass-subclass)
Only generic ’is a’ -relations are allowed
Unlike in a taxonomy, multiple inheritence is allowed.
Properties of classes
Classes can be:
Disjoint
(if n is a member of A, n is not a member of B)
(e.g. if Robin is a girl, then Robin is not a boy)
Equivalent
(if n is a member of A, n is also a member of B)
(e.g. Firstgraders Pupils born in 2003)
Exercise Create a taxonomy with these classes:
Bicycle
Boat
Bulldog
Car
Cat
Colour
Dog
Dolphin
Flower
Man
Oak
Person
Pet
Pinetree
Plant
Puppy
Rose
Whale
Woman
Zebra
Instances
Individual entities that can populate any number of classes.
An instance that is a member of a class, is necessarily also a member of all its superclasses.
Exercise Create these instances:
The semantic triple
A semantic triple is a statement consisting of three parts:
an instance
a property that refers to that instance
a value for that property
George likes chocolate
Properties
The instances are described through properties .
There are two different types of properties:
Object property:
Takes another instance as value
e.g. Alice knows Fred
Datatype property
Takes a distinct value, like a number, a string etc.
e.g. King Harald has year of birth 1937
The property is the ”predicate” in the semantic triple.
Properties of properties
Properties can be:
symmetric
(Martin has cousin Thomas) (Thomas has cousin Martin)
asymmetric
(Martin is father of Rosie) (Rosie can not be father of Martin)
inverse
(Martin is parent of Rosie) (Rosie is child of Martin)
transitive
(Rosie descends from Martin) and (Martin descends from Emma)
(Rosie descends from Emma)
functional (can have only one value)
inverse functional (value can be held by only one instance)
Domain and Range
The domain and range of a property determine what kind of instances it can be used for and what kind of values it can have.
Domain
The class, whose instances can have the property
If domain is not set, domain=Thing
Range
The class, whose instances can be value for an object property
The type of data that is allowed as value for a datatype property
Exercise Object properties
Create the following object properties
owns
hasOwner
hasNeigbour
Set domain and range correctly
Connect instances, so that:
Mr. Taylor owns Duchess
Mrs. Robertson owns Lassie
Mr. Taylor and Mrs. Robertson are neighbours
Restrictions
Classes can be populated according to rules called restrictions.
This is done by expressing that a class is equivalent to a certain set of instances.
The set can be defined by
combining other classes with and/or/not operators
using criteria based on desired properties for the instances
Examples of restrictions
LivingThing Animal or Plant or Person
Property hasOwner some Person
Pet Animal and (hasOwner some Person)
Exercise Person-Place ontology
Create a class hierarchy to describe persons by gender and function
e.g. Man, Woman, Author, Actor, Artist, Composer, Scientist, Biologist, Astronomer, Politician etc.
Create classes to describe different kinds of regions
e.g. Continent, Country, City etc.
Create object properties
isFrom-property (person-region)
partOf-property (region-region)
Create instances of both types, relate them with suitable properties
0 comments
Post a comment