A Registry for Dublin Core

1
A Registry for Dublin Core

• Thomas Baker, GMD
• IuK 2000 "Information, Knowledge
• and Knowledge Management
• Darmstadt, 27 March 2000

2
Metadata is a language

• A metadata "sentence" might say
• "This watercolor has Painter Joseph Beuys, Title
  Ohne Titel, and Date Painted 1959.
• Dublin Core was designed as a simple metadata
  language -- a"pidgin" of general concepts for
  coarse-grained resource discovery.
• In unqualified Dublin Core, the sentence above
  would say
• "This resource has Creator Joseph Beuys, Title
  Ohne Titel, and Date 1959.

3
Like languages, schemas evolve

• Like words in languages, metadata terms may be
  coined, adopted, approved by official bodies,
  change meaning, or fall from use.
• As in languages, need for simplicity is
  inevitably in tension with need for complexity
• Dublin Core Element Set is almost always too
  simple to use "as is", so it is extended locally
• If not managed, the proliferation of local
  extensions threatens interoperability in broader
  context

4
Registries as dictionaries

• Metadata systems, like languages, need
  dictionaries for tracking usage and managing
  change
• Like language dictionaries, registries can
• Prescribe good grammar and good usage guidelines
• Describe how implementors are actually using
  metadata
• Translate between natural languages
• Define the "parts of speech" of metadata grammar
  -- building blocks of sentences

5
Requirements for the DCMI registry

• Users and implementors need
• a dictionary of terms
• a place to publish project- or discipline-specific
  adaptations to share with colleagues and
  partners
• Dublin Core Metadata Initiative needs
• to manage its namespace (as a standards agency)
• to provide machine-readable schemas for loading
  into editors and search engines
• to provide crosswalks to related schemas
• to link the (English) standard to translations in
  other languages (25 to date)

6
DCMI Registry Working Group

• An RDF schema registry has already been deployed
  to support review processes within DCMI
• Working Group propose policy guidelines for
  managing the DCMI namespace with this registry
• Later encourage implementors of DC-related
  schemas (adaptations, profiles, translations) to
  put RDF schemas on the Web and link to them

7
RDF as a publication format for schemas

• How standards are are currently defined
• in HTML pages and paper documents
• no explicit hyperlinks between related elements
  in different standards
• RDF Schema format (based on XML)
• URIs provide cross-references to related schemas
  and documentation
• on Web, browse related namespaces (and profiles)
• richer thesaurus relations will support
  crosswalks between elements that are not exactly
  equivalent

8
Current prototype DCMI registry

• Based on MyRDF Toolkit
• Eric Miller, Online Computer Library Center,
  Dublin (Ohio) and RDF Working Group of W3C
• To define, search, and navigate among distributed
  collections of RDF schemas
• Works like the Web itself to add a schema, make
  it available in RDF on the Web and create
  hyperlinks to and from that schema
• Points towards a scalable ecology of metadata
  registries on the Web

9
Linking multiple translations of a standard

• Model Multilingual Dublin Core Project
• University of Library and Information Science,
  Tsukuba, Japan
• Uses RDF schemas to share machine-readable tokens
  for translations of DC terms in Japanese, Arabic,
  Punjabi (26 languages to date)
• Java applets for displaying fonts
• Raises policy questions How can we manage the
  evolution of Dublin Core as a multilingual
  standard? How can other language communities
  help shape the global (English) standard?

10
Tracking linguistic variation and equivalences

• Model MetaForm Project
• State and University Library, Goettingen
• Local "manifestations" of Dublin Core for
  specific projects introduce variations -- like
  "dialects"
• "Crosscuts" -- how are elements used in different
  implementations?
• Provides "mappings" and "crosswalks" between
  Dublin Core and other schemas of similar scope
• Demonstrates the sort of output one would want
  from queries to a distributed registry

11
Mapping between schemas via an interlingua

• Model DESIRE Project Metadata Registry
• UK Office of Library and Information Networking,
  Bath
• Maps various schemas to a core of shared concepts
  (interlingua)
• In DESIRE Registry, based on ISO Basic Semantic
  Registry
• Suggests how interoperability might be achieved
  among multiple schemas in a scalable manner
  (n-to-1 instead of n-to-n)

12
Namespaces versus Profiles

• Implementors usually need to "mix and match"
• use parts of one standard with parts of another
• coin some local terms to fill in gaps
• Application profiles (DESIRE Project)
• schemas are defined in namespaces
• namespace semantics reused in application
  profiles
• Registries should include application profiles
• expressible using RDF schema format
• will to help implementors learn from peers

13
Annotation vocabularies

• Registration authorities or third parties layer
  annotations on metadata schemas or elements
• For example, DCMI could "recommend" an element or
  qualifier -- whether it is in DCMI's own
  namespace or elsewhere
• RDF schemas support this
• Supports notion of "publish first, filter later"
  (Wilensky talk this morning)

14
DCMI Usage Committee

• Reviews elements and qualifiers in light of
  grammatical principle
• Levels of annotation (under discussion)
• local terms in use by projects
• terms proposed for review by Usage Committee
• terms found conforming to grammar principles
• terms recommended by Usage Committee
• terms that have become obsolete
• Versioning and life-cycle of terms (under
  discussion)

15
Metadata grammar

• DCMI grammar Elements and Qualifiers
• Resources (Web pages, books, museum objects) have
  things like Creators and Titles -- Elements
• Elements are modified by Qualifiers (adjectives)
• RDF grammar Resources, Relations, Classes
• RDF and DC grammars are close conceptually,
  though terminologies differ
• Using RDF schemas for the DCMI registry is
  helping clarify differences and harmonize grammars

16
thomas.baker_at_gmd.de