Making sense of schemas

1
Making sense of schemas

• Thomas Baker, Fraunhofer-Gesellschaft
• Fourth SCHEMAS Workshop
• 30 November 2001, The Hague

2
What are schemas?

• Declare, like a dictionary, vocabulary terms and
  definitions (semantics)
• Confusingly, the generic term implies both
• Describing a set of concepts and relationships
  between those concepts, and
• Encoding its machine-processable representation.
• People associate latter with W3C specifications
• RDF Schema
• XML Schema
• SCHEMAS specific uses of schema and profile
• Caveat the collective jargon is not yet stable!

3
Motivation for SCHEMAS

• Help implementors understand the evolving
  landscape of metadata schemas
• Overviews of standards development
• Lists and reviews of schema-creating activities
• Standards-based registry (in RDF)
• Harvest metadata schemas from maintainers
• Provide integrated access to this distributed
  corpus of metadata vocabulary terms
• Encourage use of existing and emerging standards

4
Where we stand

• Providing integrated access to schemas evokes a
  broader problem
• Explicitly or not, the schemas of the world
  follow a diversity of incompatible data models
• Merging diverse data models requires translation
  into a common grammar
• This talk Creating coherence (making sense)
  among diverse schemas can entail an imperfect
  process of translation, even simplification.

5
An XML schema example

• Music catalogue using an XML schema specifies a
  particular nested tag structure
• Applications that share this schema can be
  searched in a consistent manner
• User query Find identifiers of all tracks with
  creator Don Van Vliet
• Program action Find values of dcidentifier for
  track elements which have a dccreator child
  element with content Don Van Vliet

6
(No Transcript)
7
Each schema a languageunto itself

• Each XML schema defines a particular model for
  nesting tags
• Without knowing this, cannot safely guess
• Who is the creator?
• What is the relationship between a creator and a
  track?
• Web crawlers reading this metadata without its
  schema cannot make sense of it

8
Too many ways to do it!

• "XML allows users to add arbitrary structure to
  their documents but says nothing about what the
  structures mean" Tim Berners-Lee
• Different XML schemas have different structures,
  all "good" (and valid)
• Humans may be able to interpret, but machines
  need prior knowledge of parent-child element
  relations (schemas and DTDs)
• Not scalable in an open Web, where machines are
  always encountering unknown schemas

9
W3C Semantic Web

• Simple linked data model
• Create webs of information about related things
  using explicit statements
• Statements follow a common model and use
  machine-processable vocabularies
• URIs unique Web addresses for resources
• URIs tie metadata vocabulary terms to unique
  definitions that everyone can find on the Web
• XML namespaces unique addresses for metadata
  vocabulary terms
• XML universal file format

10
Semantic Web hypothesis

• A shared grammar is needed to ensure that humans
  and software will interpret metadata consistently
• A grammatical framework for the description of
  resources (Resource Description Framework)
• Clusters of simple Subject-Predicate-Object
  statements can describe most of the data
  processed by machines
• More complex grammars will not interoperate in
  the diverse Web environment

11
The RDF model

• A resource has some property whose value is
  either (i) a simple string value (literal)

http//pj.org/doc/1
author
Pete
Subject resource identified by the URI
http//pj.org/doc/1 Predicate has property
author Object the value of the property is
Pete
12
The RDF model

• or (ii) another resource...

http//pj.org/doc/1
author
name
email
Pete
pete_at_pj.org
Object of (I) is another resource Subject of
statement has name Pete and Subject of
statement has email pete_at_pj.org
13
The RDF model

• which may itself have a URI

author
http//pj.org/doc/1
http//pj.org/person/pete
name
email
Pete
pete_at_pj.org
14
URIs as anchors formerging data

• URIs are fixed points on global Web for
• Identifying resources to be described
• Identifying precisely the metadata vocabulary
  used to describe those resources
• These points can be used to superimpose graphs,
  merging statements
• Creates market for aggregation, data merging,
  annotation, and filtering services

15
First source
author
http//pj.org/doc/1
http//pj.org/person/pete
name
email
Pete
pete_at_pj.org
ltrdfRDF xmlnsuchttp//www.ukoln.ac.uk/core/gt
ltrdfDescription abouthttp//pj.org/doc/1gt
ltucauthorgt ltrdfDescription
abouthttp//pj.org/person/petegt
ltucnamegtPetelt/ucnamegt
ltucemailgtpete_at_pj.orglt/ucemailgt
lt/rdfDescription lt/ucauthorgt
lt/rdfDescriptiongt lt/rdfRDFgt
16
Second source
http//pj.org/doc/1
subject
XML
ltrdfRDF xmlnsuchttp//www.ukoln.ac.uk/core/gt
ltrdfDescription abouthttp//pj.org/doc/1gt
ltucsubjectgtXMLlt/ucauthorgt
lt/rdfDescriptiongt lt/rdfRDFgt
17
Third source
organisation
http//pj.org/person/pete
UKOLN
ltrdfRDF xmlnsuchttp//www.ukoln.ac.uk/core/gt
ltrdfDescription abouthttp//pj.org/person/pet
egt ltucorganisationgtUKOLNlt/ucorganisationgt
lt/rdfDescriptiongt lt/rdfRDFgt
18
Three descriptions merged
ltrdfRDF xmlnsuchttp//www.ukoln.ac.uk/core/gt
ltrdfDescription abouthttp//pj.org/doc/1gt
ltucauthorgt ltrdfDescription
abouthttp//pj.org/person/petegt
ltucnamegtPetelt/ucnamegt
ltucemailgtpete_at_pj.orglt/ucemailgt
ltucorganisationgtUKOLNlt/ucorganisationgt
lt/rdfDescription lt/ucauthorgt
ltucsubjectgtXMLlt/ucsubjectgt lt/rdfDescriptiongt
lt/rdfRDFgt
19
Three descriptions merged
20
Partial understanding

• To share data between programs and resources
  designed independently
• Essential trait of a massively distributed Web
• Incorporate and re-purpose data for unanticipated
  uses
• Communication among diverse communities on basis
  of partial, imperfect understanding
• Assumption tolerate inconsistency and errors!
• Ignore the ones you don't understand
• On the Web, "Error 404 File not found", but
  unchecked exponential growth

21
Pidgin metadata

• Tourists use simplified speech ("pidginisation")
  "Zwei Bier bitte".
• We are all "tourists" on a global Web with
  linguistically diverse metadata
• Core vocabulary of terms generally useful for
  description
• Dublin Core Creator, Title, Subject, Date...
• Simple Metadata Hypothesis simple metadata
  plus powerful search engines is cost-effective

22
Making schemascomparable

• SCHEMAS Expressing metadata vocabularies in a
  common grammar
• Instead of asking machines to understand people's
  language, ask people to make the extra effort
  Tim Berners-Lee
• As with natural language, translation into a
  shared grammar may involve simplification
• Allows construction of...

23
Registries

• Web-based dictionaries of metadata terms using
  machine-processable schemas
• URIs ensure that vocabulary terms (elements) are
  defined at unique locations on the Web
• Namespace schemas declare definitions for
  metadata terms ("standards")
• Application profiles mix and match standards
  for specialised needs

24
Like dictionaries

• As with natural languages
• Prescribing definitions and guidelines
• Describing actual metadata usage
• Translated definitions in French and Farsi
• Tracking evolutionary change
• Making metadata language visible, helping it
  evolve bottom-up
• Documentary and social purpose
• Define semantic coherence across applications
• Support schema design and harmonisation

25
but also unlikedictionaries

• Natural-language dictionaries
• Compiled centrally by editorial boards
• Metadata dictionaries
• Thousands of metadata schemas, dynamically
  changing
• Can only scale if vocabularies are harvested over
  the Web directly from their maintainers
• Implies shared conventions and grammars for
  publishing schemas on the Web

26
Life on the bleeding edge

• SCHEMAS has made progress on defining and
  articulating conventions for publishing metadata
  vocabularies
• much work and consensus-building
• But is RDF ready for prime-time?
• RDF specifications continue to evolve
• RDF tools are cutting-edge
• SCHEMAS registry prototype demonstrated
  proof-of-concept, but not production-ready

27
Keep it simple

• Speaking practically, broad-brush
  interoperability entails
• Partial understanding (Semantic Web)
• Shared grammar for simple statements (RDF)
• Core vocabularies, pidgin metadata (eg Dublin
  Core)
• Broadly understood conventions for publishing
  metadata vocabularies

28
Thomas.Baker_at_gmd.de