The Semantic Web
Evangelos Milios
1
Why Semantic Web?
• Web content now is mostly in natural
language with HTML annotations
• Easy for humans but difficult for computers
to interpret
• Very hard to write programs that perform
tasks for users based on Web content
• We further need
– Structured Web content
– Metadata (information about information)
– Standard content representations (shared) 2
Example
• Book an air ticket from Air Canada
• Why is the task tedious?
– All interaction is manual
– Interaction is repetitive
– Switching between screens and menus
– Shopping is difficult
• What would it take for a computer program
(“agent”) to perform the task for you?
3
Interim solutions
• Screen scraping: get the data off the
screen into a more structured form
(e.g. RDF)
• Form filling
• Things won’t get easier until many people
start publishing data in a more structured
form.
• What does this mean?
4
What is required?
• Structure: XML
– Everyone can create tags
– Says nothing about what the tags mean
• Meaning:
– Resource Description Framework
• (Subject, verb, object) triples distributed over the
Web
– Ontologies
• Taxonomies
• Logic (rules, inferences) how powerful?
– Powerful enough to be useful
– Not too powerful that makes reasoning impossible 5
RDF
(Resource Description Framework)
• Meaning is expressed as triples of:
– Things (e.g. people, objects, web pages)
– Properties (e.g. is author of)
– Values of properties (other things)
• Like subject, verb, object of a sentence
• Example: (Ullman, is author of, MySQL textbook)
• Things and Properties identified by Universal
Resource Identifiers (URI)
• Triples in RDF form webs of information about
related things 6
RDF Schema
• RDF Schema
– template for RDF documents, just like
DB schemas are templates for databases
– itself expressed in RDF
• Example:
– an RDF schema for resumes.
7
RDF example
(documents and their authors)
Application
Level
8
From http://www.w3.org/TR/2002/WD-rdf-schema-20021112/
Ontologies
• Additional vocabulary and formal semantics
compared to XML, RDF, RDF-S
• OWL: Web Ontology Language (see Wikipedia, too)
– OWL Lite:
• classification hierarchy
• simple constraints
– OWL DL:
• more expressiveness
• completeness
• decidability
– OWL Full:
• max expressiveness
• no computational guarantees
9
OWL Lite
• Class: group of individuals sharing properties
e.g. Deborah, Frank are members of class Person
• Specialization Hierarchy: subClassOf
• Property: relationship between individuals or from
individuals to data values
e.g. hasChild, isAuthor,hasAge
• Property hierarchy: subPropertyOf
e.g. hasSibling vs. hasRelative
• Domain of a property
limits individuals to which property applies
e.g. hasChild has domain Mammal
• Range of a property
limits individuals property can have as values
e.g. hasChild has range Mammal
• Individual: instance of a class
10
Equality
• equivalentClass: e.g Car , Automobile
• equivalentProperty: e.g. hasHead ,
hasLeader
• statement about individuals
– sameAs: e.g. EMilios same as EEMilios
– differentFrom:
– AllDifferent:
11
Property Characteristics
• inverseOf: e.g. hasChild, hasParent
• TransitiveProperty: e.g. hasAncestor
• SymmetricProperty: e.g. isFriend
• FunctionalProperty: e.g. hasPrimaryEmployer
(no more than one for every individual)
• InverseFunctionalProperty: inverse has
functional property
e.g. hasSIN (isSINfor)
12
Ontology examples
• Ontologies expressed in RDF syntax
• SchemaWeb: directory of RDF schemas
and OWL ontologies
– Standards ontology
– Ontology for instructional objects
• MINDSWAP Cancer ontology
• Magpie Semantic Web browser
• Friend of a Friend project
13
Automated Reasoning
• A program called reasoner can perform
inferences using the components of an OWL
ontology:
– If Frank hasChild Anna, then Frank isMammal (domain)
– If Sara isAncestorOf Louise and Louise isAncestorOf John
then Sara isAncestorOf John (transitive Property)
– No two different persons can have the same SIN
– If two instances of person have the same SIN, they are the
same person
14
Proof
• Once systems follow logic, they can prove
things
• Example:
The law says that it is a crime for an
American to sell weapons to hostile
nations. The country Nono, an enemy of
America, has some missiles, and all of its
missiles were sold to it by Colonel West,
who is American.
Prove that Colonel West is a criminal.
15
Facts and Rules
Break natural language statements into units of individual
facts and rules
16
Proof by contradiction
Combine individual facts and rules into new facts,
aiming to arrive at a contradiction.
There are often multiple ways to arrive at a contradiction
No guarantee that the search for a contradiction will terminate (undecidability)
17
First proof
8
10 and 16 are a contradiction 18
Second proof
As an exercise
19
Proofs
• Difficult to create proofs
• Easy to check them
• A Web of information processors:
– Follow rules and statements to draw conclusions
– Place results back on the Web
– Rules, statements
• distributed over the Web
• Accessible via URI’s
• “Properly designed, the Semantic Web can
assist the evolution of human knowledge as a
whole”
From: The Semantic Web: A new form of Web content that is meaningful to computers will unleash a revolution
of new possibilities, By Tim Berners-Lee, James Hendler and Ora Lassila, Scientific American, May 17, 2001
20
Summary
• Semantic Web
• RDF
• RDF schemas
• OWL
• OWL Lite
• Reasoning
21