Dean Allemang Semantic Web Basics

TopQuadrant Semantic Web Technology Training
Series

TopQuadrant in collaboration with Jim Hendler
presents:
“Getting Ready for the Semantic Web with
TopBraid Suite”
Module I-2:
Overview of Semantic Technologies and the
Semantic Web

What is Semantic Technology?
What is it good for?

The Semantic Wave is NOT one thing … there are
differing major streams within it

The Semantic Web
W3C Standards for sharing information on a world-wide
scale
Intranets vs. Internet

Semantic Technology
Enhanced knowledge access and search
Semantic Interoperability
Information syndication
… and so forth

© Copyright 2007-2009 TopQuadrant Inc. Slide 2

Philosophy of Semantics
What is Semantics = “meaning of meaning”
Logical Positivism (with a vengeance!) =
“Everything is described by values of properties”
Shirky: “Semantic Web is about tautologies.
(…and tautologies aren’t interesting)”
Intelligent Agents = AI on the web (it’s gotta be
smart!)


Semantic Web: Make web content machine-readable!

“The Semantic Web is a vision: the idea of having data on the Web defined
and linked in a way that it can be used by machines not just for display
purposes, but for automation, integration and reuse of data across various
applications.[W3C 2001] ”

“The Semantic Web is an extension of the current Web in which information is
given well-defined meaning, better enabling computers and people to work in
cooperation.” [Tim Berners-Lee et al 2001]


What could the Web do?

Web page interaction –
uses people as its medium!


What could the Web do? (cont.)

Can this sort of
interaction
become part of
the Web itself?


How could the Web do it?

Built-in by the Webmaster

Agree upon an “interlingua”


What about XML? Doesn’t it support semantics?

HTML gave us formatting tags
XML gave us custom tags
You get to pick your tags/attributes
Tags can have “meaning” specific to your application

Many dialects have blossomed
XML and XML Schema became W3C standards
Standard dialects are being developed by many
industry groups – XBRL.org, FpML.org,
TaxML.org, …
Every large organization has their own XML
Schemas

Gartner: All Tied Up with XML: 1999

Unprecedented growth of standard development

Gartner: All Tied Up with XML: 2001

From 2001 through 2004 enterprises spent $3 billion on modeling activities with no
return on investment from $ 2 billion of it.

A new Web of terminology

What’s the Interlingua for the Interlingua?

Use the same technology for mapping web pages
to terminology
to map terminology to one another

What about people? Don’t they make the world go
round?

computers and people…better cooperation

Internet
Ontology
Ontology
Ontology

Annotated Agent Human
Annotated
Web Page
Annotated
Web Page
Web Page

Source:, Phil Windridge WSWS (2004)

The Web: The World’s Largest
Information System!

How did it get so big? What is special about The Web?


How the Web was Won – The Network Effect

The “Network Effect” –
The more participants,
the more value in joining

At first, there were few web pages
Little value in joining in.

But then there were more –
more value in joining in

A crowd is better yet!
Lots of value in joining in


Features of The Web
Anyone can say Anything about Any topic (AAA)
All names are global (so that anyone can refer to them!)

Two people might have different names for the
same thing . . . (non-unique naming)
. . . Or the same name for different things!

You never know everything on the web (“Open
World”)

This isn’t what we want the web to be, it is how the
web is (and how it supports the network effect!)

AAA Slogan -

Anyone

can say
Anything

about Any topic


Non-unique naming

“Java”
public String getContextPath() {
try {
Method getContextPathMethod = “Java”?
delegate.getClass().getMethod("getContextPath", null); //$NON-NLS-
1$
return (String) getContextPathMethod.invoke(delegate, null);
} catch (Exception e) {
// ignore “Coffee”?
}
return null;
}

Hot Beverage?
Programming language?


Open World Assumption
There’s always something else over the horizon . . .


Web of Ontologies

Network Effect for
Ontologies:

The more there are, the
easier and more valuable
it is to create a new one

From Tim Berners-Lee, ISWC 2003

Semantic Solutions -
Expectations we have encountered

A solution is "Semantic" if it:
X understands the meaning of natural language
X interprets perceptual input and forms usable representations
X computes new statistical analyses of data
X recognizes complex patterns in huge amounts of data
X produces high-quality graphics of complex systems
X replaces an intelligent human in a high-impact task
X has a model of uncertainty in data
X can come to correct conclusions despite faulty input.

Companies provide (some of) these, but
typically as “add-ons” or 3rd party solutions
Not as part of core Semantic offerings


W3C standards for semantic models
W3C Semantic stack is built on XML
XML-based Ontology languages are being
developed to support semantic interoperability.

“Semantic Web is stimulating a
whole new class of applications
at individual, enterprise and web
scales”
– Eric Miller, W3C, Semantic
Technologies for eGOV’2003
www.w3.org/
www.w3.org/2001/sw


W3C standards evolution
“Layer cake” has been modified a few times…

Significant changes:
•Addition of a query language
(SPARQL)
•RDF not dependent on XML
•OWL just one of many possible
logic formalisms

Not shown: emergence of n3
serialization (built for triples, unlike
XML)

Series

presents:
TopBraid Suite”
Module I-3:
Semantic Web

Mapping the Semantic Terrain:
Standards and Languages from
the W3C

The Tree of Knowledge Technologies

Content
Management
Semantic Languages
Technology
Languages
Process
Knowledge
Languages

AI Knowledge Software
Representation Modeling
Languages


Semantic Web Standards Stack


How Semantic Languages Work
Bring information together
Draw inferences

RDF OWL

RDFS


What is RDF? Distribution of data
Model Manufacture In
ID Division Product Line SKU
No. location Stock
Manufacturing
1 ZX-3 Paper machine Sacramento FB3524 23
support
Manufacturing
2 ZX-3P Paper machine Sacramento KD5243 4
support
Manufacturing
3 ZX-3S Paper machine Sacramento IL4028 34
support
Control
4 B-1430 Feedback Line Elizabeth KS4520 23
Engineering
Control
5 B-1430X Feedback Line Elizabeth CL5934 14
Engineering
Control
6 B-1431 Active Sensor Seoul KK3945 0
Engineering

7 DBB-12 Accessories Monitor Hong Kong ND5520 100

8 SP-1234 Safety Safety Valve Cleveland HI4554 4

9 SPX-1234 Safety Safety Valve Cleveland OP5333 14


Distribute by rows?
Manufacturing Paper
1 ZX-3 Sacramento FB3524 23
support machine

Needs common schema Control
- which column is which? B- Feedba Elizabet KS452
4 Engineerin 23
1430 ck Line h 0
g

7 DBB-12 Accessories Monitor Hong ND5520 100
Kong


Distribute by columns?

Model No.

ZX-3

ZX-3P Needs to reference
ZX-3S entities – which thing are Division

B-1430
we talking about? Manufacturing
support
B-1430X Manufacturing
support
B-1431
Manufacturing
DBB-12 In Stock
support
SP-1234 23 Control
Engineering
SPX-1234 4
Control
34
Engineering
23
Control
14 Engineering

0 Accessories

100 Safety

4 Safety

14


Distribute by cells!?

Division

7 Accessories

Needs to reference both
Division schema and entities
7 Accessorie
Most flexible – can distribute
s
data in any way at all!
Product
Line
4 Feedback
Line

Model
1 ZX-3


Distribute by cells!?

Predicate
Division

7 Accessories

Subject

Object

•Store,
•Index, and
•Federate
these triples
URI’s
Triple Store


Representing Data in Graphs
Graph = nodes linked by labeled edges


What is RDF?
RDF (Resource Description Framework) is an
infrastructure for:
Encoding,
Exchange and
Distributing metadata
A “Resource“ is anything that you want to describe
RDF Triple:
RDF Triple:

Subject Object Object
Subject
Predicate
Predicate


Why is RDF useful?
After all, can’t we express semantics in XML?

XML allows application-specific tags
Semantic exchange can happen, as long as two parties agree on the use of
these tags

XML Represents a tree structure – no indication
about what is a description, what denotes an
object, what values relate to what objects, etc.

RDF provides a way to integrate structured
information from multiple sources


What is RDFS?
RDFS is the schema language for RDF
Type inferences can be made, based on schema


Why is RDFS useful?
RDFS allows us to talk about classes of instances

It provides inferences, e.g.,

Best Western is a Hotel

(and hence, anything we know about Hotels
applies to Best Western)

RDFS is in RDF (it’s its own schema language!)


How does it work?

RDF – the Ultimate Mash-up
Language !!

RDF

RDF

RDF
RDF

A little RDF(S) goes a long way

gov: body
eGovOS: project

gov: department

eGOV: web
gov: agency service

eGOV:Service Spec
eGOV: Remote
Reporting
gov: DoE gov: FERC

gov: EPA

eGOV: capability
brm: s2citizens brm:Energy brm:Resource eGOV:
Mgmt Standard

brm:
brm: Business Area brm: Line Of
subfunction
Business

A model of government agencies and
departments. Such models are called
Ontologies.

Ontologies are the means to separate “what is
common” from “what is different”

Semantic map:
Connecting silo
domains

From Tim Berners-Lee, ISWC 2003

OWL (formerly DAML+OIL)

What is OWL?
DARPA EU (various)

DAML OIL

The “Web RDF

Ontology
Language” DAML+OIL

W3C Standard

OWL W3C

Became a
Recommendation in
February 2004

OWL can specify rich relationships: equivalence,
inverse, unique, …


What is OWL good for?
OWL provides a flexible way to talk about sets of
resources,
e.g., “All planets around the sun”, “The wives of Henry
VIII”, “People who have visited Japan”, “Incidents of SIDS
in which the mother is related to someone with epilepsy”
A powerful way to relate information
“people who have flown to Japan have been to Japan”, “a
person with epilepsy has a neurological disorder”,
“planets are astronomical bodies”
Provides a robust way to re-use information
Even if I didn’t expect someone to be interested in the
planets around the sun when I built my database, you can
define this set and draw conclusions about it (e.g., “there
are 9 of them”)

SPARQL

SPARQL Protocol and RDF Query Language

Query Language (like SQL is for databases,
XQuery is for XML, etc.)

Extracts information from a graph using pattern
matching

“Four-star lodging in New York”


Find data in a graph


Series

presents:
TopBraid Suite”
Module I-5:
Semantic Web

Comparing Semantic
with
Conventional Technologies

Semantic Technology, RDBMs and Object Models

In this session we will compare Semantic
Technology with
Relational Database and
Object Oriented Technology
We will use as an example a simple application for
managing and tracking medical equipment
application will be asked to find an address
compare semantic and relational database approaches
We will then identify key differences and
similarities between semantic and object models


Semantic and OO Technologies

Comparing Semantic Technology (ST)

and

Object Oriented (OO) Technology


Differing Intents of Semantic and Object
Models
Object Model
A specification of how a set of entities can encapsulate
data and invoke behaviors on one another
The intent of the model is to provide realizable software
where object behaviors become fragments of code

Semantic Model (Ontology)
A specification of what is known in a region of interest
Intent is to maintain consistency between what is known
about a domain in general (expertise) and what is
observed about a situation (data)


Specification (OO) vs. Knowledge Discovery (ST)

OO: In an Object Model class membership and
hierarchy are specified
Classes are defined and express ‘constraints’ on the instances
(individuals) that belong to classes

ST: An Ontology model can serve either as a
specification of class membership or as a means of
knowledge discovery
Sets of properties associated with individuals (instances) may
allow them to be viewed as members of several classes at the
same time
Class membership is dynamic and depends on the value of the
properties
In OWL, classes are computed based on assertions about the
© Copyrightindividuals that might belong to one or more classes
2007-2009 TopQuadrant Inc. Slide 5

(ST) In OWL, Classes are inferred or computed

OWL classes are interpreted as sets that contain
individuals
A class is not a kind of template as in OO technology
In OWL, classes are built up of descriptions that specify
the conditions that must be satisfied by an individual to be
a member of the class

Subclasses are subsets of their parent classes.

Superclass-subclass relationships can be
computed automatically by a reasoner


OO Systems Define Acceptable Expectations for
Instances of Classes and Subclasses

Information Given
Medical Equipment Class What can you do?
method list_tests
PW1729.list_tests(..)

X-Ray Machine Class This call must be valid.
extends Medical Equipment (and is done like it is for all
Medical Equipment, unless
stated otherwise)

Instances
Instance Class Relationships in the model
PW1729 X-Ray Machine determine what can be done,
and what it means to do it.


Properties (ST) vs. Attributes and Relations (OO)

OWL Properties represent relations between two
individuals (Not classes)

OWL Property types:
Object Properties link an individual to an individual

Datatype Properties link an individual to simple values
• integers, floats, strings, booleans, and so forth
• an XML Schema Datatype primitive value or an RDF literal

Those with OO experience/expertise must
overcome the typical pre-conception that
properties belong to the class!


Properties are first-class constructs
This allows relationships between Properties
In contrast to most OO paradigms, where properties are
“owned” or “contained in” Classes

hasParent
This allows relationships
between Properties
hasMother hasFather

This is not a class diagram!

dc:creator

… and for other modelers to
my:author reuse properties
“wherever I use the property
‘author’, tell the world that
they can read ‘dc:creator’”


In OWL, Properties may have Sub Properties

It is possible to form hierarchies of properties
(these are not Class hierarchies)

The rdfs:subPropertyOf construct allows
relationships to be abstracted up the sub-property
tree.

Properties may have Additional Characteristics

OWL allows the meaning of properties to be enriched
through the use of property descriptions:
Sub-property: Relation between properties, that the pairs related by
one property are included in the other.

Domain and Range: Descriptions of a property that determines class
membership of individuals related by that property.

Inverse: Description of a property that exchanges the Subject and
Object in their respective relationship in another property that it is an
inverse-of.

Transitive: Chains of relationships collapse into a single relationship.

Symmetric: Description of a property that makes it a self-inverse
property

Functional and Inverse Functional

…

What does it mean to be a subclass in OWL?

In OWL, subclass means necessary implication

If A is a subclass of B, then ALL instances of A are instances of B,
without exception.

If something is an A, then this implies that it is also a B

OWL allows the use of class expressions in place of named
classes
restrictions describe an unnamed set that could contain some
individuals
these are anonymous, computed classes

Properties are used to define and describe Classes
Properties are used to create restrictions.
Restrictions are used to restrict the individuals that belong to a class.
When you define a restriction, it is the property that give rise to the
class

Class Expressions and (Multiple) Inheritance

Semantic Web Classes are (logically) sets, so you
can do set logic on them:

Person

Parent
Female Male

Mother ∩

“A ‘Mother’ is a ‘Female’ who is also a ‘Parent’”


(ST) Run-time vs. (OO) Design time semantics

In semantic systems, ontologies are run-time
models

There is no separation between the model and its
technical implementation

The model (ontology) is the OWL code

Semantic web technology Model-driven
applications
vs.
OO technology Model-developed applications

OO Development – Progressive Design and
Transformation of Classes into Code

Requirements Analysis Objects Design Objects

Concepts
Use Cases Class Code

(ST) No behavior is
VOPOS or CRC CRC
described anywhere –
only inferencing

Scenarios Method Code

Expected Object Responsibilities Object Services Method Code
Behavior (public interface) (signatures)

Natural Language Refactoring Write code that fulfills
Contract Validation Add Signatures the defined signatures,
Object Discovery Adapt to Constraints Adding private/helper
Human Effort Pick Mechanisms Methods as needed


Summary Table: Understanding the “Lingo”

OO
Object Technology Semantic Technology ST Term
Term
Class A “Factory for instances” Definition of a set of Class
Definition of behavior individuals. Uses open world
(methods, values). approach – defines some, but
Defines all the attributes not necessarily all possible
and associations. properties of its members.
Instance Member of one direct Member of one or more Individual or
class, responds to all classes, as determined by the Instance
methods defined in the instance property values and
class class definition
Attribute Data encapsulated with Data describing an individual DatatypeProperty
an instance. Defined only (or instance). Defined globally.
for the members of the
Association class
Reference pointer from Reference from one RDF ObjectProperty
one object instance to resource to another. Defined
another. Defined only for globally.
the members of the class
Inheritance Methods and properties Domains and ranges inherits Inference
inherit down up. OWL constraints inherit
© Copyright 2007-2009 TopQuadrant Inc. down. Slide 16

Semantic Web – OO Gotchas!

In the Semantic Web, you infer the class of an object.
The class of an object can change:
over time
with what you know/believe
with whom you trust
Properties are first-class objects (independent of classes!)
Properties form hierarchies as well as classes
No behavior is described anywhere – only inferencing
Multiple set membership is commonplace
No OO inheritance


OWL / RDFS Preview (for Comparison)

owl:inverseOf ‘child inverseOf parent’
Elizabeth has child Charles,
therefore Charles has parent Elizabeth

owl:transitiveProperty
Elizabeth has descendant Charles,
Charles has descendant Andrew,
therefore Elizabeth has descendant Andrew

rdfs:subPropertyOf ‘child subPropertyOf
descendant’
Elizabeth has child Charles,
therefore has descendant Charles

Semantic vs Relational

Comparing Semantic Technology

and

Relational Database Technology


Data in a spreadsheet

Person Company

Melli Annamalai Oracle

Xavier Lopez Oracle

Dean Allemang TopQuadrant

Mike Uschold Boeing

Ora Lassila Nokia


Same data in a relational database

Give each entity a
number (a ‘key’) so you Express
Person Table
have an unambiguous relationships by
way to talk about it. associating these
ID Name
numbers together
in another table.
1 Melli Annamalai Company Table

Works For ID Name
2 Xavier Lopez
1 1 1 Oracle
3 Dean Allemang 2 1 2 TopQuadrant
3 2
4 Mike Uschold 3 Boeing
4 3
4 Nokia
5 4
5 Ora Lassila


Same data in a relational database

Works For Company Table
Person Table
1 1 ID Name
ID Name
2 1 1 Oracle
1 Melli Annamalai 3 2 2 TopQuadrant
4 3 3 Boeing
2 Xavier Lopez 5 4 4 Nokia

3 Dean Allemang Country
Based in
ID Name
4 Mike Uschold 1 1
1 USA
2 1
2 Finland
5 Ora Lassila 3 1
4 2


Suppose (as a simplification) that we know that
people live in the country where their employer is
based.

Query:SELECT all the people and the countries where
List p.name, s.name
they FROM employment.person p,
live Person p works for
company c
employment.company c,
employment.country s,
employment.basedIn bi, It is the ‘person’ field of the
employment.worksFor wf worksFor table that
WHERE corresponds to the person
wf.person = p.id AND It is the ‘company’ field of
wf.company = c.id AND the worksFor table that
bi.company = c.id AND corresponds to the
bi.country = s.id ; company

Same data in a graph
Melli Annimalai
Oracle

Xavier Lopez USA

TopQuadrant

Dean Allemang

Finland

Boeing
Mike Uschold

label
worksFor (person, company)
basedIn (company, country)
Ora Lassila Nokia


Suppose (as a simplification) that we know that
people live in the country where their employer is
based.

Query:SELECT all the?sname and the countries where
List ?pname people
they WHERE
live Person p works for
company c
{ ?p :worksFor ?c.
?c :basedIn ?s.
?p rdfs:label ?pname .
?s rdfs?label ?sname .}
It is the ‘person’ field of the
The ?p that :worksFor
worksFor table that
something is a Person
corresponds to the person
(in the schema)


RDFS and OWL are about Rich Relationships in Data

Information Given
X-Ray Machine located in Room A2003
Room A2003 located in ATA Building
Information Inferred
Where are the
relationships? X-Ray Machine located in ATA
Building
Relationship Model
Question
“located in” is a transitiveProperty Where is the X-Ray
Machine located?
Relationships are explicit in the Answer
model and directly available to
applications! In ATA Building


OWL has many Build-in Functions for modeling
Relationships

Information Given
X-Ray Machine stored in Room A2003
Room A2003 located in ATA Building Information Inferred
X-Ray Machine located in Room 2003
X-Ray Machine located in ATA Building

Relationship Model Question
Where is the X-Ray Machine
“located in” is a transitiveProperty located?
“stored in” is a subPropertyOf “located in”
Answer
Also inverse, functional, local In ATA Building
contextual constraints, … Still works!


Getting the same result with a relational database

Equipment Table

ID Equipment Name
Question
Relationships are in
Where is the X-Ray
IDQ X-Ray Machine documents and in
Machine located?
collective memories -
Building Table not available to
applications! Develop a Query
Room Building
A2003 ATA Building SELECT Building
FROM Equipment Table,
Building Table,
Room_Equipment Table Room_Equipment Table
WHERE Equipment Name = “X-
Room_ID EQ_ID Data Definition Statements?
Applications do not use them, they are Ray Machine” and ID = EQ_ID
not descriptive and their scope is a
A2003 IDQ single database
and Room = Room_ID

Data Dictionary? Data Registry?
Answer
Where are the They are for human, not
computer use In ATA Building
relationships?


Change happens…

Information Given
Room A2003 located in ATA Building
X-Ray Machine located in Room A2003
Information Inferred
Room A2003 part of Dr Smith’s Office
Room A2003 located in
Dr. Smith’s Office located in ATA Building
Dr. Smith’s Office
Room A2003 located in ATA
Building
X-Ray Machine located in ATA
Building
Relationship Model
Question
“located in” is a transitiveProperty Where is the X-Ray
“part of” is a subPropertyOf “located in” Machine located?

Answer
new
In ATA Building


Accommodating change with RDB requires database
changes AND new queries

Equipment Table
ID Equipment Name Equipment Table
IDQ X-Ray Machine
ID Equipment Name
IDQ X-Ray Machine
Building Table
Business Name ID Building Building Table
Dr Smith MDS ATA Building
Room Building
A2003 ATA Building
Room_Equipment Table
Room ID Equipment ID
A2003 IDQ Room_Equipment Table
Room_ID EQ_ID
Room_Business Table A2003 IDQ

Room_ID Bus_ID
A2003 MDS


Accommodating change with RDB requires database
changes AND new queries

Equipment Table
ID Equipment Name
IDQ X-Ray Machine Question
Where is the X-Ray
Machine located?
Building Table
Business Name ID Building Using the same query:
Dr Smith MDS ATA Building

SELECT Building
Room_Equipment Table
FROM Equipment Table,
Room ID Equipment ID Building Table,
A2003 IDQ Room_Equipment Table
WHERE Equipment Name = “X-
Room_Business Table Ray Machine” and ID = EQ_ID
and Room = Room_ID
Room_ID Bus_ID Answer
A2003 MDS
?

Comparing two approaches
RDB Semantic Model (Ontology)
Database must be designed Ontology must be designed to
Ability to to answer the questions answer the questions
Answer
Specific, typically complex, Queries can be generic and
Questions
queries must be developed very simple
Inflexible: Flexible:
Ability to Database structure must be Ontology can be easily extended
modified so it can continue so it can continue to answer new
Accommoda to answer the questions questions
te Change Queries must be re-written No data porting required
Data must be ported

Can be very fast with proper Not as fast, but improving,
tuning – mature technology: tuning does not affect
Known optimization flexibility:
Processing approaches Adding more processing
Speed Certain queries, such as multi power and distributed
computation helps
table joins and self joins are
known to cause problems Performs better than RDBMS
for certain query types

Key differences in the representation of relationships

RDB Semantic Model (Ontology)
Relationships are either 1:1, By default all relationships are
many:1 or many:many many:many
Many:many relationships must be Functional properties and
Cardinality of broken into many:1 relationships cardinality restrictions are used to
relationships by creating join tables specify 1:1, many:1 as well as other
cardinalities
It is possible to specify, for example, 1:4
or min 2, etc.

Additional information about the Relationship is reified (made into a
Information relationship is represented by the class)
bearing extra columns in the join table Additional information is
relationships represented as properties of the
class
Implicit Explicit
Embedded in the name of the Care is taken to name a relationship
join table or in the name of the in a way that its nature and
The nature of column intentions are well understood
relationships Typically these names are not
designed for ease of
understanding of the nature of
© Copyright 2007-2009 TopQuadrantthe relationship
Inc. Slide 33

Another Key Capability of Semantic Web Technology is
Managing Distributed Data

Information does not have to be in a single data
source
Information about equipment can be in one place
Information about buildings, addresses and doctor
offices in a different place
RDF provides and infrastructure for merging and
unifying data in a consistent way


Relational and Semantic Technology can work well
together: it is not necessarily one or the other!

Semantic Application
Interaction Logic
Application Logic
Semantic Interface

WS
Semantic Hub Schema Enterprise Query
Translation Ontology Ontology
Models Models Models

Mapping Mapping Mapping Mapping

Legacy

WS
WS
WS

WS
Systems
SI SI SI SI
IL IL
ERP Data
AL AL Warehouse

BL CRM BL PLM


TopQuadrant Semantic Web Technology Training Series

TopQuadrant in collaboration with Jim Hendler presents:
“Getting Ready for the Semantic Web with TopBraid
Suite”

Module IIa-4:
Using Semantic Standards, Languages and
TopBraid Tools for Modeling and Querying

Querying RDF with SPARQL

Query Languages
Familiar query languages:
SQL
XQuery

SPARQL is similar to these in that it allows one to
specify patterns of data

SPARQL differs in syntax, e.g., no notion of “JOIN”


Extracting information from RDF
RDF is a graph structure
How do I get information from it?
Where is Stratford?
Who married the person who wrote King Lear??
Did the person who wrote Hamlet live in a Hamlet?


Querying some sample data
Table: Play
ID Title Written By Year
1 The Tempest Shakespeare 1611
2 Romeo and Juliet Shakespeare 1595
3 As You Like It Shakespeare 1599
4 Edward II Marlowe 1592
5 Dido, Queen of Carthage Marlowe 1586
6 Eastward Ho Johnson 1605
7 A Game at Chess Middleton 1624
8 Sir Thomas More Munday 1592
The Tragical History of
9 Marlowe 1604
Doctor Faustus


Same data viewed as a graph
Play
…
The Tempest Row1 Shakespeare

Romeo and Juliet Row 2 Marlowe

As You Like It Row 3 Johnson

Edward II Row 4 Middleton

Dido Row 5 Munday

Eastward Ho Row 6
1586 writtenBy
1592
Game at Chess Row 7 1595 year
1599
Sir Thomas More Row 8 title
1604
1605 rdf:type
Dr. Faustus Row 9 1611
1624


Titles of plays written by Shakespeare

The Tempest Row 1 Shakespeare




Dido Row 5 Munday

Eastward Ho Row 6
1586 writtenBy
1592
1599
Sir Thomas More Row 8 title
1604
1605
1624


Matching data with graph patterns

The Tempest Row1 Shakespeare

Romeo and Juliet Row 2

As You Like It Row 3

?x ?y Shakesepeare

writtenBy
?x ?y
How do we know that year
The Tempest Row 1 these things are plays?
title
Romeo and Juliet Row 2 What is a better heading
for column “?x”?
As You Like It Row 3


A more complex graph pattern





Dido Row 5 Munday

Eastward Ho Row 6
1586 writtenBy
1592 year
Game at Chess Row 7 1595
1599 title
Sir Thomas More Row 8 1604
1605
1624
?x ?z ?a

The Tempest
Tempest Shakespeare 1611

?x ?y ?z As You Like It
Like Shakespeare 1599

Dido
Dido Marlowe
Marlowe 1586
?a Thomas More Munday 1592

© Copyright 2007-2009 TopQuadrant Inc. etc. Slide 8

Representing graph patterns
Graph patterns are represented just like graphs
– in triples!
Variables allowed as well as resources:
?x ?y :Shakespeare

?y :title ?x.
?y :writtenBy :Shakespeare .

?title ?play :Shakespeare

?play :title ?title. writtenBy
?play :writtenBy :Shakespeare . year

title


More graph patterns

?x ?y ?z

?a

writtenBy

year
?y :title ?x.
?y :writtenBy ?z. title
?y :year ?a.

?title ?y ?playwright

?year

writtenBy

?y :title ?title. year
?y :writtenBy ?playwright.
title
?y :year ?year.

More graph patterns
“The names of two playwrights who wrote plays in the same
year”
Play

?play2 ?playwright2
?playwright1 ?play1

?year

?play1 rdf:type :Play .
writtenBy
?play1 :writtenBy ?playwright1 . year
?play2 :writtenBy ?playwright2 .
title
?play1 :year ?year .
?play2 :year ?year . rdf:type

This finds all plays, so need some way to state that ?playwright1
is different than ?playwright2 (FILTER)

Building Queries from Data
“The names of two playwrights who wrote plays in the same year”

Play

Sir Thomas More Munday
Marlowe Edward II

1592

writtenBy
?play1 :writtenBy ?playwright1 . year
title
?play2 :year ?year . This is the basis of the rdf:type
automatic query builder in
TopBraid Composer and
Ensemble


A more complex graph pattern
Play
…




Dido Row 5 Munday

Eastward Ho Row 6
1586 writtenBy
1592
1599 title
Sir Thomas More Row 8 1604
1605 rdf:type
1624

Play

?play2 ?playwright2
?playwright1 ?play1

?playwright1 ?playwright2
?year
Marlowe Munday

Components of a Query
Selection mode and variables
SELECT ?place
SELECT ?playwright1 ?playwright2
CONSTRUCT {?playwright rdf:type :Person}

Triple patterns – like a triple, but with named variables
instead of some parts
?play :title ?title .
?play :playwright :Shakespeare .
?play :year ?year .
SPARQL standard is written similar to N3
Filters ?play2 rdf:type :Play .
numerical comparisons, calculations
?year > 1600 ?play1 :year ?year .
© Copyright 2007-2009 TopQuadrant Inc.
FILTER (?playwright1 != Slide 14

Query Syntax
Bring it all together to form a real query:

SELECT ?place
WHERE { :Stratford :isIn ?place .}
Where is Stratford?
SELECT ?spouse
WHERE
{ ?spouse :married ?author . Who married the person who
?author :wrote :KingLear .} wrote King Lear?

SELECT ?place
WHERE Did the person who wrote
{ ?author :wrote :Hamlet . Hamlet live in a Hamlet?
?author :livedIn ?place .
?place rdf:type geo:Hamlet . }
Separates graph triples
Best practice: end
every triple with a '.'

Filters
“Shakespearean plays written after 1600”

You can only filter on values you have matched in the WHERE clause!

First, find Shakespearean plays and the years they were written:

writtenBy
?title ?play Shakespeare year
title

?year

SELECT ?title ?year
WHERE { ?play :title ?title . ?title ?year
?play :writtenBy :Shakespeare . The Tempest 1611
?play :year ?year .
Romeo and Juliet 1595
FILTER (?year > 1600) As You Like It 1599
}


Optional Patterns
“Titles and authors of all plays”
Suppose we have some plays whose author is unknown:

ID title writtenBy year
101 Maid’s metamporphosis 1600
102 Revenger’s Tragedy 1607
4 Edward II Marlowe 1592

Must match… Optional match
writtenBy
?title ?play ?author year
title
SELECT ?title ?author
WHERE {?play :title ?title . ?title ?author
OPTIONAL {?play :writtenBy ?author .} Maid’s metamporphosis
}
Revenger’s Tragedy
Edward II Marlowe

Negation
“Titles of anonymous plays”
ID title writtenBy year
101 Maid’s metamporphosis 1600
102 Revenger’s Tragedy 1607

writtenBy
?title ?play ?author year
title
Filter out those for which the author was found using the filter function
“bound”:
The variable ?author was ?title
SELECT ?title
not bound to a value
Maid’s metamporphosis
WHERE {?play :title ?title .
Revenger’s Tragedy
OPTIONAL {?play :writtenBy ?author .}
FILTER (!bound (?author))
}


Searching for Literals
Previous queries matched a variable (e.g. ?title)
find something with a specific literal (xsd) value
Who wrote “A Game at
SELECT ?author Chess”?
Order matters: each graph
WHERE { ?play :title "A Game at Chess" .
pattern further restricts the
?play :writtenBy ?author .
} overall match

SELECT ?author
WHERE { ?play :title "A Game at Chess"^^xsd:string .
?play :writtenBy ?author .
} specify xsd type

SELECT ?author
WHERE {?play :title ?title . Regular expression defined by XQUERY
FILTER (regex(?title, "a game", "i")) . 1.0
} regex(?title, "[a-zA-Z]{2}[aeiou]s{2}$”)
matches when end of string is “Chess” (as
© Copyright 2007-2009 TopQuadrant Inc. well as other string combinations) 19
Slide

Union
Find all plays written by
Shakespeare either in 1611 or 1595
SELECT ?play ?title ?year
WHERE
{
{ ?play :writtenBy :Shakespeare .
?play :title ?title .
?play :year "1611"^^xsd:string . The query matches two different graph patterns:
} 1.Matches ?play and ?title with specified :year
UNION 2.Matches ?play and ?year filtered by ?year
{ ?play :writtenBy :Shakespeare ; ?play has different binding in each of the graph
:year ?year . patterns
FILTER(regex(?year, "1595")) .
} ?play ?title ?yea
}
:Play1 The Tempest r
:Play2 1595


Query Forms
SELECT
returns matches for specified variables
CONSTRUCT
returns a graph that includes triples constructed from specified
variables
ASK
true if a match to the graph pattern is found
DESCRIBE
returns a graph (instead of selecting “columns”) – not supported
by TBC


ASK
ASK returns a Boolean value
based on whether WHERE clause finds data

Were any plays written after 1600?

Most useful in programs
Result displayed in status bar
SPIN
SPARQLMotion, JSP, etc.


CONSTRUCT
SPARQL queries can build new sets of triples based on
patterns
WHERE clause exactly as before
CONSTRUCT returns a triple pattern
Subtly powerful:
Under these conditions, add these triples
Basically what reasoners do (datalog)

plays are written by playwrights

CONSTRUCT {?person rdf:type :playwright}
WHERE { ?play :writtenBy ?person .
?play rdf:type :Play .}


Constructing new triples with SPARQL

Play 1 Shakespeare

Play 2
Marlowe
Play 3
Johnson Playwright
Play 4
Middleton
Play Play 5
Munday
Play 6 writtenBy

rdf:type
Play 7

Play 8

Play 9

SELECT CONSTRUCT
Play ?pla ?person ?person Playwright
y Same binding re-used
from body to head

TopQuadrant Semantic Web Technology Training Series

TopQuadrant in collaboration with Jim Hendler presents:
“Getting Ready for the Semantic Web with TopBraid
Suite”

Module IIa-3:
Using Semantic Standards, Languages and
TopBraid Tools for Modeling and Querying

SKOS – Simple Knowledge
Organization System

What is a Knowledge Organization System?

A means of organizing and communicating
terminology about a domain
Controlled Vocabulary
Taxonomy
Thesaurus

Thesaurus standards
NISO Z39.19, ISO 2788-1986(E), ISO 5964-1985(E)
Provide for relationships between terms
• BT, NT, RT
• Preferred terms, alternate terms
Monolingual, Multilingual, hierarchical


Vocabulary Uses

Tagging, e.g.:
del.icio.us, Flickr
West Key Number System
Library of Congress etc.
Entity Extraction, e.g.:
Calais
Allow one set of expert users (‘catalogers’) to inform another set (information
customers) to communicate about the content of information items
(documents, photos, web pages, etc.)

Data Modeling
Data dictionaries, schemas
Spreadsheet columns
Supports consistency in data modeling, promoting interoperability.


Agreeing on terms

Why can’t everyone just
agree to use terms the same
way?

… Everyone has their own
business processes and
stakes.

Disagreement is legitimate!

Simple Knowledge Organization System

SKOS is the RDF standard for Knowledge Organization
Systems
Allows different groups to organize terminology…
… while allowing them to link to one another

“The St. Louis notion of ‘Customer’ is broader than the
New York notion of ‘Customer’”

You can’t even have this discussion, if you don’t recognize
that there are actually two contexts for the word
“Customer”!

Mapping terms

broader match

“Customer”
“Customer”


SKOS Resources

Qname label Abbreviation
(where applicable)
skos:broader “has broader term” BT
skos:narrower “has narrower term” NT
skos:related “has related term” RT
skos:broadMatch “has broader match”
skos:narrowMatch “has narrower match”

skos:prefLabel “preferred label”
skos:altLabel “alternative label”


Dean Allemang Semantic Web Basics

Recommended

Recommended

More Related Content

What's hot

What's hot (8)

Viewers also liked

Viewers also liked (20)

Similar to Dean Allemang Semantic Web Basics

Similar to Dean Allemang Semantic Web Basics (20)

Recently uploaded

Recently uploaded (20)

Dean Allemang Semantic Web Basics