Guidelines for machineprocessable representation of DCAPs

1
WI 4 (CWA1) Guidelines for machine-processable
representation of Dublin Core Application Profiles
Pete Johnston, UKOLN, University of Bath Thomas
Baker, Fraunhofer-Gesellschaft CEN/ISSS MMI-DC
Meeting Brussels, 22-23 September 2004
http//www.ukoln.ac.uk/
2
Machine-processable representation of Dublin Core
Application Profiles

• Context
• Conceptual model for DCAP
• Suggested representation using RDF

3
Context

• Metadata application profile
• Recognition that implementers adapt metadata
  standards to context
• Use terms from multiple metadata vocabularies in
  combination
• CEN CWA 14855
• Guidelines for human-readable representation of
  DCAP
• Current draft
• Make information available in structured form,
  usable by applications
• Influenced by
• DCMI practice ("Grammatical Principles,
  Namespace Policy, declaration of metadata
  vocabularies, DCMI Abstract Model)
• W3C Semantic Web activity
• Research projects on metadata schema registries

4
DCMI Abstract Model

• Working Draft of DC Architecture WG
• Seeks to make explicit the DC "meta-model"
• what are the component parts of any DC metadata
  description
• what information these components convey about
  the resources described by the DC metadata
  description
• independent of form in which DC metadata
  description is represented
• closely aligned with RDF meta-model
• adopts class hierarchy/property specialisation
  semantics of RDFS

5
DCMI Abstract Model

• Description as set of statements about a subject
  resource
• Each statement describes a relationship between
  the subject resource and a second resource
  (value)

Ref to resource
Statement
Ref to property
Ref to value
Description
6
Fundamentals of DCAP

• DCAP does not define new terms
• DCAP references ("uses") terms already defined
  elsewhere
• Terms may be from multiple independently-created
  sources
• DCAP may describe how use of terms is
  constrained, adapted, contextualised
• N.B. CWA 14855 employed "term usage" current doc
  employs "property usage
• DC differentiates different types of term
• Only use of properties is constrained
• Other types of term are referenced
• Only (or at least primarily) as part of
  constraints on property

7
Conceptual model for DCAP

• What is a DCAP?
• What are the component entities? What are the
  related entities?
• What are the attributes of a DCAP? And of these
  component and related entities?
• What types of relationship exist between these
  entities?

8
1
administers
1
isDescribedIn
1
m
DCAP
Agency
SchemaDocument
1
1
1
hasPropertyUsage
isExpressedBy
administers
m
m
m
1
m
1
m
usesProperty
hasTerm
BindingSchema
Property
MetadataVocabulary
PropertyUsage
1
1
m
isDescribedIn
hasTerm
usesAsEncodingScheme
1
n
m
Class
SchemaDocument
9
MetadataVocabulary
10
Metadata Vocabulary

• A set of metadata terms (Properties, Classes, and
  Instances of those classes) managed as a coherent
  unit by an Agency

Examples the DCMES, the DC Terms Vocabulary,
the DCMI Type Vocabulary
11
Property

• A Property is a type of relationship between two
  Resources.
• A Property is declared as a term within exactly
  one Metadata Vocabulary.
• A Property may be related to another property by
  a sub-property relationship this states that all
  resources related by the first property are also
  related by the second property

Examples dccreator, dctermsmodified,
dctermsaudience (All DCMI elements and element
refinements are properties.)
12
Class

• A Class is a group of resources.
• A Class is declared as a term within exactly one
  Metadata Vocabulary.
• A Class may be related to another class by a
  sub-class relationship this states that all
  instances of the first Class are also instances
  of the second Class.
• A Resource is related to one or more Classes by a
  type relationship, and is said to be an Instance
  of those classes .

Examples dctermsLCSH, dctermsW3CDTF,
dcmitypeText, dcmitypeCollection (All DCMI
"encoding schemes" and type vocabulary terms are
classes.)
13
DCAP
14
DC Application Profile (DCAP)

• A set of Property Usages, created to meet the
  functional requirements of an application or
  context, and managed as a coherent unit by an
  Agency.

Examples the Simple Dublin Core DCAP, the
RDN-DC DCAP, the Renardus DCAP
15
Attributes of DCAP
16
Property Usage

• A Property Usage is a description of how a
  previously declared Property from a Metadata
  Vocabulary is deployed in the context of an
  application.
• A Property Usage
• must reference ("use") exactly one Property.
• may provide additional documentation on how the
  property is interpreted in the context of this
  application
• may provide an application-specific label for the
  property
• may specify obligation for the use of statements
  referring to the property (whether it is
  mandatory, optional, conditional)
• may specify constraints on the occurrence of
  statements referring to the property
• may specify constraints on the permitted values
  of the property, by specifying that they are
  instances of specified classes (i.e. may specify
  "encoding schemes" for the property)

17
Property Usage
Examples the usage of dctitle in the Simple
Dublin Core DCAP, the usage of dctitle in RDN-DC
DCAP, the usage of dctitle in Renardus DCAP
18
Attributes of Property Usage
19
Attributes of Property Usage
20
Representation of DCAP XML?

• Could provide a XML DTD or XML Schema to define
  an XML format for a DCAP
• But the property usages in a DCAP reference
  existing terms
• Term descriptions already available, using
  RDF/RDFS (in some cases, at least!)
• Would require
• re-describing terms that are already described
  (or map existing data to new format) or
• using separate format/model for DCAP (DCAP-XML)
  and for metadata vocabulary (RDFS/RDF)

21
Representation of DCAP XML Schema?

• An XML Schema describes constraints on the
  structure of a (class of) XML document(s)
• Abstract Model
• Description may be represented as records in
  multiple syntaxes
• May be multiple XML formats, each with different
  XML Schema
• A DCAP specifies the properties/classes used in a
  description
• So (potentially) one-to-many relation between
  DCAP and XML Schema

22
Representation of DCAP XML Schema?

• However, XML implementers want
• to constrain structure of DC-in-XML documents
  during creation
• to validate structure of DC-in-XML documents
  post-creation
• so need XML Schema corresponding to DCAP (for
  their chosen XML format)
• DCAP model is (probably?!) rich enough to
  generate XML Schema
• but N.B. that generation process requires
  additional information about each XML format

23
Representation of DCAP RDF?

• RDF provides simple meta-model
• Resource-property-value
• Descriptions of terms in DCMI metadata
  vocabularies already published using RDF
• using RDFS and DC vocabularies
• Many other significant vocabularies also
  available currently or will be available
• By definition DCAP references other terms
• Use of RDF facilitates merging of DCAP
  description and existing metadata vocabulary
  descriptions (and resource descriptions)

24
Representation of DCAP RDF?

• However, DCAP concept is closely associated with
  that of document/record/bounded description
• mandating that statement with specified property
  is present
• limiting number of occurrences of statements with
  specified property
• mandating that value of specified property is
  instance of specified class
• Generally, RDF applications tend to adopt
  "open-world" assumptions
• RDFS, OWL designed to support inferencing, rather
  than completeness/correctness checks (validation)

25
RDF representation

• Specify RDF classes and properties corresponding
  to the entity types, attributes, relation types
  in model
• Use existing RDF vocabularies where possible
• RDF Vocabulary Description Language (RDF Schema)
  provides
• a semantics of class hierarchy/property
  specialisation
• an RDF vocabulary to represent RDFS semantics
• i.e. properties and classes to describe
  Properties, Classes (and Datatypes)
• DCMES/DC terms provide
• properties for many descriptive attributes

26
RDF representation

• RDFS has no concepts of application profile,
  property usage
• RDFS does not provide
• a class to represent a (Metadata) Vocabulary
• So need to provide additional classes and
  properties where required
• The dcap vocabulary
• Should provide RDFS descriptions of dcap terms
• N.B. No URIrefs yet assigned for dcap terms

27
Example

• RDN-DC
• DCAP used for record-sharing between partners in
  Resource Discovery Network (RDN)
• Sharing over OAI-PMH, so uses XML syntax
• Usage of dclanguage
• Optional (recommended)
• Repeatable
• Requires use of RFC3066 encoding scheme

28
(No Transcript)
29
ltdcapPropertyUsagegt ltdcapuses
rdfresource"dcnslanguage"/gt
ltdcdescriptiongtUse the language codes defined in
RFC 3066.lt/dcdescriptiongt ltdcapobligation
rdfresource"dcapnsObligation/recommended"/gt
ltdcapmaxOccursgtUnboundedlt/dcapmaxOccursgt
ltdcapencodingScheme rdfresource"dctermsnsRFC3
066" /gt ltdcapisMemberOf rdfresource"http//
www.rdn.ac.uk/ap/rdn_dc"/gt lt/dcapPropertyUsagegt
30
Issues

• Choice of URIrefs for dcap RDF vocabulary terms
• Currently, no DCMI-endorsed model for DCAP
• Proposed model is largely untested!
• But JISC IEMSR registry in development (similar
  data model)
• DCMI Abstract Model still work-in-progress
• Literal and non-literal values in DC metadata?
• Use of literal datatyping for syntax encoding
  schemes?
• DCAP for description v DCAP for description set
• CEN CWA 14855
• more "permissive" view of DCAP?