Postcards from the Semantic Web' Metadata for Machines

1
Postcards from the Semantic Web. Metadata for
Machines

• INF 384C
• Miles Efron
• School of Information
• University of Texas

2
Metadata for Humans and for Machines

• We can understand the role of metadata in modern
  information access along several axes.
• Metadata type (descriptive, structural, admin)
• Who is the metadata for?
• How informative is the structure latent in the
  metadata?
• Todays lecture has two goals
• To contextualize our earlier discussion of
  metadata use in the real world.
• To articulate an important current in the
  metadata world creating metadata for use by
  machines instead of humans (grandly exemplified
  by the Semantic Web initiatives)

3
Debating the merits of the Semantic Web

• How do the proponents of Semantic Web
  technologies and initiatives argue that SW is
  different (and better) than the Web as we know
  it?
• Why does the Semantic Web have so many, and such
  vocal detractors?

4
Metadata and the Semantic Web

• The semantic web is a conglomeration of projects
  that share the desire to create and maintain a
  distributed information system based on highly
  structured metadata.
• The operation of the semantic web is predicated
  on the existence of shareable ontologies.
• The lingua franca of these ontologies is the
  resource description framework (RDF).

5
The Idea(l) of the Semantic Web

• Instead of a Web of pages intended for human
  interpretation, the Semantic Web consists of
  linked pages intended for interpretation by
  semi-autonomous software agents.
• Thus instead of the browser serving as your
  gateway to Web-based info, you might have, for
  example, a Calendar agent that helps you
  negotiate people, places, and times.

6
The Idea(l) of the Semantic Web
For this to work, such an agent would need to
know not only that people, places, and times
exist, but how they relate to each other, and how
they are structured. e.g. A boss is a kind of
person, and so is an assistant. Also, two people
can occupy one place, but a person can only
occupy one place at a time.

• Instead of a Web of pages intended for human
  interpretation, the Semantic Web consists of
  linked pages intended for interpretation by
  semi-autonomous software agents.
• Thus instead of the browser serving as your
  gateway to Web-based info, you might have, for
  example, a Calendar agent that helps you
  negotiate people, places, and times.

7
What do we talk about when we talk about an
Ontology?
short answer An ontology is a specification of
a conceptualization.
more generally an ontology is a description ...
of the concepts and relationships that can exist
for an agent or a community of agents. T.
Gruber. What is an Ontology?
8
In less lofty terms Ontology
An ontology is a representation of the things in
a system, represented in a consistent,
agreed-upon fashion.
Typically, the term ontology implies not only a
representation of things, but also a
representation of their relationships, along with
semantic understanding of those
relationships. i.e. a relational database may or
may not be an ontology, depending on your
definition.
9
Ontologies aid in Knowledge Sharing
knowledge about physics
knowledge about chemistry
how to compare/combine knowledge that is
represented in different ways?
10
Ontologies aid in Knowledge Sharing
knowledge about physics
knowledge about chemistry
analogous representations allow for direct
comparisons
11
Ontologies aid in Knowledge Sharing
knowledge about physics
knowledge about chemistry
analogous representations also encourage
combining knowledge bases
12
Things you can do with an Ontology

• Issue queries

• which uranium isotopes are most prone to chain
  reaction?
• what's a nice red wine to eat with meunster?
• show me all the people who have siblings named
  dave.

• Make Assertions

• rich people make make at least 100,000
• it is not the case that I make more than
  100,000
• therefore I am not a rich person

the point of building an ontology is to enable
processing of knowledge by other programs...the
ontology itself is rarely the goal.
13
Elements of an Ontology
English
Ontology-speak

• concepts
• features
• actual things

• classes (e.g. wine, car)
• slots (e.g. type, cost)
• instances

make
car
model
year
14
Elements of an Ontology
name
is it a class or a slot???
country
make
car
model
year
15
Elements of an Ontology
Classes are the things that comprise the little
universe described by an ontology.
These classes are represented primarily by slots,
i.e. variables. Slots may themselves be classes.
16
Organization of an Ontology
Ontologies typically encode associative
relatioinships between classes.
car
is-a relation
sedan
17
Organization of an Ontology
Ontologies typically encode associative
relatioinships between classes.
bird
has-a relation...i.e. meronymy
beak
18
Organization of an Ontology
The notion of inheritance adds efficiency to
representation.
weight
make
thing
model
car
sedan
superclasses vs. sub-classes
coupe
19
Expressing Ontologies with XML

• The core idea underlying semantic web efforts is
  that there should exist numerous ontologies, all
  of which should be expressed in a way that allows
  them to be shared/combined/used by a variety of
  systems.
• The semantic web community has focused on using
  XML as the lingua franca for expressing
  ontologies.
• In particular, several XML markup languages
  underpin the use of ontologies for semantic web
  activity

20
Expressing Ontologies with XML

• RDF The Resource Description Framework. Used
  to express simple assertions about the
  relationships between things.
• RDF Schema A language used to describe the
  relationship between classes of things, or
  relationships between concepts in a particular
  domain.
• OWL The Web Ontology Language. Relies on both
  RDF and RDF schema to express the ontologies
  themselves.

21
The Resource Description Framework (RDF)

• http//www.w3.org/RDF

describes
anything
Many of the following examples are adapted from
http//www.w3.org/TR/rdf-primer and
http//www.dlib.org/dlib/may98/miller/05miller.htm
l .
22
RDF Overview

• When we talk about RDF, we commonly refer to at
  least three distinct ideas
• RDF is a conceptual structure, a model for
  expressing descriptive statements about
  resources.
• RDF is often expressed (aka "serialized") using
  XML.
• RDF/XML is informed by ontological knowledge,
  expressed using another XML-based language RDF
  Schema.

23
Problem DTDs specify syntax, not semantics
lt!ELEMENT letter (greeting,body,closing)gt lt!ELEMEN
T greeting (salutation?,recipient)gt lt!ELEMENT
body (PCDATA)gt lt!ELEMENT closing
(signoff?,sender)gt lt!ELEMENT salutation
(PCDATA)gt lt!ELEMENT recipient (PCDATA)gt lt!ELEMEN
T signoff (PCDATA)gt lt!ELEMENT sender
(PCDATA)gt lt!ATTLIST letter letterDate CDATA
REQUIREDgt
what is a sender? what is a sender's
relationship to a letter?
24
Problem DTDs specify syntax, not semantics
In fact, it can be argued that DTDs do provide
semantics. i.e. Dublin Core offers an author
element, with fairly obvious meaning. But what
we lack in this circumstance is a way to compare
semantics across DTDs. RDF seeks to remedy this.
The goal of RDF is to promote semantic
interoperability, much as DC promotes syntactic
interoperability.
25
The RDF Model Thinking of Metadata in terms of
Semantics
http//www.ibiblio.org/mefron/ has a creator with
name Miles Efron
http//www.ibiblio.org/mefron/ has a creation
date with value August 1998
http//www.ibiblio.org/mefron/ has a language
with value English
26
Thinking of Metadata in terms of Semantics
resource
atomic value
property
http//www.ibiblio.org/mefron/ has a creation
date with value August 1998
27
Thinking of Metadata in terms of Semantics
resource
atomic value
property
http//www.ibiblio.org/mefron/ has a creation
date with value August 1998
RDF data types

• a resource is something that has a URI
• a property is something possessed by a resource
• an atomic value is a basic piece of data (such
  as a number)

28
RDF as a Language a sample Statement
http//www.ibiblio.org/mefron/ has a creation
date with value August 1998
subject
predicate
object
29
RDF as a Language a sample Statement
http//www.ibiblio.org/mefron/ has a creation
date with value August 1998
subject
predicate
object
Together, these are called an RDF triple.
30
RDF as a Directed Graph
resource
http//www.ibiblio.org/mefron
property
creation date
atomic value
August 1998
31
RDF as a Directed Graph
resource
http//www.ibiblio.org/mefron
property
author
resource
miles_at_utexas.edu
32
RDF as a Directed Graph
resource
Miles Efron
email
properties
name
affiliation
miles_at_utexas.edu
Efron, Miles
www.utexas.edu
33
Serializing RDF RDF/XML

• To make our RDF conceptual models
  machine-readable, the RDF working group at W3C
  has defined an XML vocabulary for describing
  resources.
• Unlike many of the other XML standards we've
  discussed, RDF/XML uses shallow, repetitive
  statements to articulate descriptions. This has
  implications for the kind of info RDF is good at
  expressing

34
Structure of an RDF/XML document
root element is RDF
the RDF element contains 1 or more Description
elements. Each Description has a resource's URI
as its about attribute
the Description element contains 1 or more child
nodes, each of which contains a predicate-value
pair.

35
Structure of an RDF/XML document
A Description element is not strictly necessary,
but they will be present in all of our RDF/XML
documents.
root element is RDF
the RDF element contains 1 or more Description
elements. Each Description has a resource's URI
as its about attribute
the Description element contains 1 or more child
nodes, each of which contains a predicate-value
pair.

36
Structure of an RDF/XML document
RDF
Description about http//www.ibiblio.org/mefron
author Miles Efron
title Miles Efron's web page
37
Properties are given semantics
RDF uses XML namespaces to specify inter-resource
relations
DCCreator
Miles Efron
http//www.ibiblio.org/mefron
ltRDFRDF xmlnsRDF"http//www.w3.org/1999
/02/22-rdf-syntax-ns" xmlnsDC"http//dubl
incore.org/2003/03/24/dces"gt ltRDFDescription
RDFabout "http//www.ibiblio.org/mefron"gt
ltDCCreatorgtMiles Efronlt/DCCreatorgt
lt/RDFDescriptiongt lt/RDFRDFgt
38
Properties are given Semantics
lt?xml version"1.0"?gt ltrdfRDF xmlnsrdf"http//w
ww.w3.org/1999/02/22-rdf-syntax-ns"
xmlnsDC"http//www.dublincore.org/2003/03/24/dce
s"gt ltrdfdescription about"http//www.ibiblio.
org/mefron"gt ltDCcreatorgtMiles
EfronltDCcreatorgt ltDCtitlegtMiles Efron's web
pagelt/DCtitlegt ltrdfdescriptiongt ltrdfRDFgt
39
Some Example RDF/XML Docs
lt?xml version"1.0"?gt ltrdfRDF
xmlnsrdf"http//www.w3.org/1999/02/22-rdf-syntax
-ns" xmlnsdc"http//dublincore.org/2003/0
3/24/dces"gt ltrdfDescription rdfabout"http//w
ww.example.org/index.html"gt ltdcdategtAugust
16, 1999lt/dcdategt lt/rdfDescriptiongt
ltrdfDescription rdfabout"http//www.example.org
/index.html"gt ltdclanguagegtenlt/dclanguagegt
lt/rdfDescriptiongt lt/rdfRDFgt
40
Expressing Dublin Core in RDF/XML
lt?xml version"1.0"?gt ltrdfRDF
xmlnsrdf"http//www.w3.org/1999/02/22-rdf-syntax
-ns" xmlnsdc"http//dublincore.org/2003/0
3/24/dces"gt ltrdfDescription rdfabout"http//w
ww.ukoln.ac.uk/metadata/resources/dc/datamodel/
WD-dc-rdf/"gt ltdctitlegt Guidance on
expressing the Dublin Core within the Resource
Description Framework (RDF) lt/dctitlegt
ltdccreatorgt Eric Miller lt/dccreatorgt
ltdcsubjectgt Dublin Core Resource Description
Framework RDF eXtensible Markup Language
XML lt/dcsubjectgt ltdcpublishergt Dublin Core
Metadata Initiative lt/dcpublishergt
ltdccontributorgt Dublin Core Data Model Working
Group lt/dccontributorgt ltdcdategt 1999-07-01
lt/dcdategt ltdcformatgt text/html
lt/dcformatgt ltdclanguagegt en lt/dclanguagegt
lt/rdfDescriptiongt lt/rdfRDFgt
Adapted from http//www.ukoln.ac.uk/metadata/resou
rces/dc/datamodel/WD-dc-rdf
41
Expressing Dublin Core in RDF/XML
lt?xml version"1.0"?gt ltrdfRDF
xmlnsrdf"http//www.w3.org/1999/02/22-rdf-syntax
-ns" xmlnsdc"http//dublincore.org/2003/0
3/24/dces"gt ltrdfDescription rdfabout"http//w
ww.ukoln.ac.uk/metadata/resources/dc/datamodel/
WD-dc-rdf/"gt ltdctitlegt Guidance on
expressing the Dublin Core within the Resource
Description Framework (RDF) lt/dctitlegt
ltdccreatorgt Eric Miller lt/dccreatorgt
ltdcsubjectgt Dublin Core Resource Description
Framework RDF eXtensible Markup Language
XML lt/dcsubjectgt ltdcpublishergt Dublin Core
Metadata Initiative lt/dcpublishergt
ltdccontributorgt Dublin Core Data Model Working
Group lt/dccontributorgt ltdcdategt 1999-07-01
lt/dcdategt ltdcformatgt text/html
lt/dcformatgt ltdclanguagegt en lt/dclanguagegt
lt/rdfDescriptiongt lt/rdfRDFgt
So what exactly do these namespace URI's point
to? What is a title, or a subject? Ideally,
these terms should be drawn from an ontology,
expressed in RDF Schema or a related language.
Adapted from http//www.ukoln.ac.uk/metadata/resou
rces/dc/datamodel/WD-dc-rdf
42
RDF Schema

• RDF gives us a syntax for expressing
  relationships between things in the world.
• But it doesnt explicitly lend these
  relationships any meaning.
• To enable systems to share meaningful, structured
  information (i.e., sharing knowledge), RDF
  statements often adhere to ontologies expressed
  in RDF Schema.

43
RDF Schema

• In our earlier discussion of XML, we talked about
  DTDs and Schemas.
• These tools define a metadata vocabulary.
• Similarly, RDF Schemas define the vocabularies
  for ontologies. Additionally, RDF Schema
  contains functions to define ontological
  relationships among an ontology's classes and
  properties (slots).

44
A simple Schema Bread

• Imagine that we are going to create an electronic
  database of bread recipes.
• The database will be expressed using RDF so that
  we can share information with other services.
• To enable that sharing, we need to define the
  kinds of knowledge that our database contains.
  Hence the schema

45
Outline of the Ontology
Has a
bread
shape
Has an
ingredient
Is a
leaven
46
Outline of the Ontology
Has a
bread
shape
Has an
ingredient
Our ovals will become classes, and the arrows
will become properties. N.B. This is Not a formal
RDF graph.
Is a
leaven
47
Writing the Ontology
lt?xml version"1.0" ?gt ltrdfRDF
xmlnsrdf'http//www.w3.org/1999/02/22-rdf-syntax
-ns' xmlnsrdfs'http//www.w3.o
rg/2000/01/rdf-schema' gt lt/rdfRDF
gt
RDF Schema documents are valid RDF, so we start
as usual, adding a namespace for the Schema
language (this gives us access to the ontology
functions).
48
Writing the Ontology
lt?xml version"1.0" ?gt ltrdfRDF
xmlnsrdf'http//www.w3.org/1999/02/22-rdf-syntax
-ns' xmlnsrdfs'http//www.w3.o
rg/2000/01/rdf-schema' gt lt!-- top-level classes
--gt ltrdfsClass rdfID'Shape' /gt ltrdfsClass
rdfID'Ingredient' /gt ltrdfsClass rdfID'Bread'
/gt lt/rdfRDFgt
49
Writing the Ontology
lt?xml version"1.0" ?gt ltrdfRDF
xmlnsrdf'http//www.w3.org/1999/02/22-rdf-syntax
-ns' xmlnsrdfs'http//www.w3.o
rg/2000/01/rdf-schema' gt lt!-- top-level classes
--gt ltrdfsClass rdfID'Shape' /gt ltrdfsClass
rdfID'Ingredient' /gt ltrdfsClass rdfID'Bread'
/gt lt!-- sub-classes --gt ltrdfsClass
rdfID'Leaven' gt ltrdfssubClassOf
rdfresource'Ingredient' /gt lt/rdfsClassgt lt/
rdfRDFgt
50
Writing the Ontology
lt?xml version"1.0" ?gt ltrdfRDF
xmlnsrdf'http//www.w3.org/1999/02/22-rdf-syntax
-ns' xmlnsrdfs'http//www.w3.o
rg/2000/01/rdf-schema' gt class declarations
omitted due to space lt!-- properties
--gt ltrdfProperty rdfID'breadName'gt
ltrdfsdomain rdfresource'Bread'
/gt lt/rdfPropertygt ltrdfProperty
rdfID'breadShape' gt ltrdfsdomain
rdfresource'Bread' /gt ltrdfsrange
rdfresource'Shape' /gt lt/rdfPropertygt ltrdfPro
perty rdfID'breadIngredient' gt ltrdfsdomain
rdfresource'Bread' /gt ltrdfsrange
rdfresource'Ingredient' /gt lt/rdfPropertygt ltrd
fProperty rdfID'breadHydration' gt
ltrdfsdomain rdfresource'Bread'
/gt lt/rdfPropertygt lt/rdfRDFgt
The ontology is available at http//www.ibiblio.o
rg/mefron/xml/rdf/bread.rdf
51
Using the ontology

• Imagine that we have a recipe for the wonderful
  Italian bread called pugliese.
• The recipe is stored at the URI
  http//www.bread.com/recipe1.html
• We want to create an entry in our database that
  describes this recipe using the knowledge we've
  just formalized.

52
lt?xml version"1.0" ?gt ltrdfRDF
xmlnsrdf"http//www.w3.org/1999/02/22-rdf-syntax
-ns" xmlnsrdfs"http//www.w3.org/2000
/01/rdf-schema" xmlnsbread"http//www
.ibiblio.org/mefron/xml/rdf/bread"
gt lt/rdfRDFgt
We declare a namespace for the ontology,
attaching it to a copy of the Schema (this is
where I've put it online).
53
lt?xml version"1.0" ?gt ltrdfRDF
xmlnsrdf"http//www.w3.org/1999/02/22-rdf-syntax
-ns" xmlnsrdfs"http//www.w3.org/2000
/01/rdf-schema" xmlnsbread"http//www
.ibiblio.org/mefron/xml/rdf/bread"
gt ltrdfDescription rdfID"batard"
rdftype"http//www.ibiblio.org/mefron/xml/rdf/br
eadShape" gt lt/rdfDescriptiongt ltrdfDescription
rdfID"durumFlour" rdftype"http//www.ibib
lio.org/mefron/xml/rdf/breadIngredient"
gt lt/rdfDescriptiongt ltrdfDescription
rdfID"wheatFflour" rdftype"http//www.ibibli
o.org/mefron/xml/rdf/breadIngredient"
gt lt/rdfDescriptiongt ltrdfDescription
rdfID"salt" rdftype"http//www.ibiblio.org/m
efron/xml/rdf/breadIngredient"
gt lt/rdfDescriptiongt ltrdfDescription
rdfID"water" rdftype"http//www.ibiblio.org/
mefron/xml/rdf/breadIngredient"
gt lt/rdfDescriptiongt ltrdfDescription
rdfID"instantYeast" rdftype"http//www.ibibl
io.org/mefron/xml/rdf/breadLeaven"
gt lt/rdfDescriptiongt
54
lt?xml version"1.0" ?gt ltrdfRDF
xmlnsrdf"http//www.w3.org/1999/02/22-rdf-syntax
-ns" xmlnsrdfs"http//www.w3.org/2000
/01/rdf-schema" xmlnsbread"http//www
.ibiblio.org/mefron/xml/rdf/bread" gt
instances omitted for space ltrdfDescription
about"http//www.bread.com/recipe1.html"
rdftype"http//www.ibiblio.org/mefron/xml/rdf/br
eadBread" gt ltbreadbreadNamegtpuglieselt/breadbr
eadNamegt ltbreadbreadShapegtbatardlt/breadbreadSh
apegt ltbreadbreadIngredientgtdurumFlourlt/breadbr
eadIngredientgt ltbreadbreadIngredientgtwheatFlour
lt/breadbreadIngredientgt ltbreadbreadIngredientgt
saltlt/breadbreadIngredientgt ltbreadbreadLeavengt
instantYeast"lt/breadbreadLeavengt
ltbreadbreadHydrationgt0.78lt/breadbreadHydrationgt
lt/rdfDescriptiongt lt/rdfRDFgt
55
Again Why the Skepticism?
56
Again Why the Skepticism?

• Some critics have argued that the Semantic Web
  has been slow to catch on because it hinges on
  persuading data owners to structure their
  information manually, often in the absence of a
  clear economic incentive for doing so. While the
  Semantic Web approach may work well for targeted
  vertical applications where there is a built-in
  economic incentive to support expensive markup
  work (such as biomedical information), such a
  labor-intensive platform will never scale to the
  Web as a whole.
• Rather than relying on Web site owners to mark up
  their data, couldnt search engines simply do it
  for them?
• Google now sends a spider to pull up individual
  query forms and indexes the contents of the form
  for clues about the topic it covers.
• --Wright, Alex. (2008). Searching the Deep Web.
  Communications of the ACM. 51(10). 14-15.

57
Again Why the Skepticism?
People sometimes talk about the semantic web (vs.
Semantic Web). Whats the difference?

• Some critics have argued that the Semantic Web
  has been slow to catch on because it hinges on
  persuading data owners to structure their
  information manually, often in the absence of a
  clear economic incentive for doing so. While the
  Semantic Web approach may work well for targeted
  vertical applications where there is a built-in
  economic incentive to support expensive markup
  work (such as biomedical information), such a
  labor-intensive platform will never scale to the
  Web as a whole.
• Rather than relying on Web site owners to mark up
  their data, couldnt search engines simply do it
  for them?
• Google now sends a spider to pull up individual
  query forms and indexes the contents of the form
  for clues about the topic it covers.
• --Wright, Alex. (2008). Searching the Deep Web.
  Communications of the ACM. 51(10). 14-15.

58
Metadata and the Semantic Web

• As weve already discussed, the idea behind the
  Semantic Web is to put in place metadata that is
  intended to help machines learn about things in a
  way that is useful to people.
• To guide this learning, stakeholders in the
  Semantic Web use a host of XML-based tools (RDF,
  OWL--the Web Ontology Language--, etc.) to encode
  ontologies, expressions of knowledge about the
  world that machines can use.
• Lets look at some Semantic Web Examples

59
Postcards from the Semantic Web

• http//www.semantic-systems-biology.org/biogatewa
  y/querying
• W3Cs RDF Validator (http//www.w3.org/RDF/Validat
  or/) Validates RDF and will generate a graph
  visualization of the marked up relationships.