CS 502: Computing Methods for Digital Libraries

1
CS 502 Computing Methods for Digital Libraries

• Lecture 18
• Descriptive Metadata Metadata Models

2
Administration
3
Metadata Models

• Challenge
• Are there general principles of metadata?
• Applicable to wide varieties of formats, genres
  and types.
• Useful for information discovery, collection
  management, preservation, e-commerce.
• Applicable to different creations, versions,
  etc.
• Balance cost against utility

4
Perceived Problems with Dublin Core
Separate metadata elements for Creator,
Contributor, and Publisher. Confusion between
Genre and Format. Many qualifiers have been
proposed for Date. Apparent redundancy of the
element "Source". Reasons for the 11
relationship between metadata and an information
resource.
5
Creations
Work
Expression
IsRealizedThrough
Manifestation
IsEmbodiedIn
Item
IsExemplifiedBy
6
Agents
Agent
or
IsOwnedBy
IsProducedBy
IsRealizedBy
IsCreatedBy
7
Actions
Transforms
PlaysRoleIn
Agent
Actions
8
Actions

• Originate
• Compile
• Excerpt
• Reformat
• Use
• Acquire/relinquish object
• Acquire/relinquish rights

9
Extras
Creations may have relations between each
other Actions may have time or place Creations
may have subjects
10
Relations
References / Is Referenced By to point to other
information resources IsBasedOn / IsBasisFor
to express intellectual derivation IsVersionOf
/ HasVersion to express historical evolution Is
Format Of / Has Format to identify
transformations of media or layout Is Part of /
Has Part to record Part/Whole
11
Relations
Is IsMetadataAuthorOf / HasMetadataAuthoredBy to
name the creator of the metadata IsDefinitionOf /
IsDefinedBy to point to the URI of the
definition of the semantics IsOwnerOf/
IsOwnedBy to name the owner/repository with
custody of a physical thing
12
Resolution of Problems with Dublin Core
Separate metadata elements for Creator,
Contributor, and Publisher. All are Agents
related to Works, Expressions, Manifestations or
Items. Confusion between Genre and Format. Both
are Form, related to either Work/Expression or
Manifestation/Item. Many qualifiers have been
proposed for Date. A Work, its Expression, a
Manifestation and an Item can each have a
particular Date. Apparent redundancy of the
element "Source". Source is expressed more
clearly as a particular Relation.
13
Unresolved problems 1

• Complex objects
• Indecs view is confused between
• single package of material for e-commerce
• trails of rights and payments of indefinite
  complexity
• Dublin Core view is limited to
• document-like objects with simple relationships

14
Unresolved problems 2

• Cost
• No system provides useful guidelines to balance
  cost of metadata against utility
• Indecs model of e-commerce requires complex
  metadata that is not needed for information
  discovery and other applications
• Dublin Core is struggling with the tension
  between minimalists and structuralists

15
Unresolved problems 3

• Documentation and training
• Both Indecs and Dublin Core will require a
  detail set of instructions and codes.
• Despite many years of effort they are a long
  way from having such guidelines.
• In a rapidly changing world, the guidelines
  will always be out of date.
• The guidelines will need trained people to
  understand them. The dream of metadata generated
  by untrained creators and publishers is far away.

16
Personal Opinion

• Metadata is valuable only because it enables
  specific services
• Information discovery (descriptive)
• Rendering and display (structural and technical)
• Commerce (administrative)
• Long-term preservation (descriptive, structural
  and technical)
• All metadata depends on its context
• "Author" is
• descriptive metadata for information discovery
• administrative metadata for commerce

17
Personal Opinion

• Metadata is expensive
• Costs must be balanced against value for specific
  services
• There is no correct place and time to gather
  metadata
• At source v. by cataloguer or indexer
• By professional v. automatically
• At time of creation v. when needed