Professional Documents
Culture Documents
Interoperability
Pete Johnston, Eduserv Foundation & Rosemary Russell, UKOLN
1 Introduction
The JISC Repositories & Preservation Programme has funded a number of projects to
work on metadata application profiles for the description of a range of resources. The
metadata application profiles developed so far - the Scholarly Works Application
Profile (SWAP)1, the Images Application Profile (IAP)2, the Geospatial Application
Profile (GAP)3 - are all Dublin Core Application Profiles, i.e. they are based explicitly
on the Dublin Core Abstract Model. The project currently working on the profile for
Time-Based Media (TBMAP) is also operating on this basis.
The SWAP, the IAP and the TBMAP each have as their focus the description of a
particular class or genre of resources. The GAP differs slightly in that it is intended to
be used in conjunction with other profiles; and it focuses on a specific set of
characteristics which may be applied to resources of many different types, the
distinguishing characteristic being that they have some relationship with “place”.
This note provides some issues for discussion around the possible uses of the profiles.
It should be emphasised that it is an attempt to raise some tentative questions, rather
than to provide definitive solutions.
1
• A specification of the structural constraints on the description sets, used to
represent instances of that domain model, i.e. a specification of the resources that
may be described, the properties referenced in statements, and the way value
surrogates are provided
• A set of guidelines for how to apply the profile
• A specification of any concrete syntax(es) to be used
i.e within the Dublin Core metadata framework, the choice of domain model and the
choice of vocabulary are addressed at the level of the DC Application Profile. It is
important to note that there is no “global” domain model provided by DCMI to
underpin all DC metadata. In particular, “Simple Dublin Core” is just one DCAP,
based on a (not always clearly articulated) “domain model” in which all resources are
treated as having the same set of 15 attributes. The same set of properties used within
the Simple Dublin Core DCAP (i.e. the Dublin Core Metadata Element Set) may be
deployed in whole or in part in the context of other DCAPs based on other domain
models.
It is perhaps also worth emphasising that two different DCAPs might be based on the
same domain model, or on variants of a single domain model, but differ in terms of
the sets of metadata terms referenced and/or the structural constraints imposed by the
description set profile.
2
3.1.2 Linking based on Multiple DCAPs
Now consider the case of two “repository” services exposing metadata based on two
different DCAPs, based on two different domain models. There may be perfectly
good reasons for those differences, based on the requirements the DCAP designers set
out to address. Although in principle, the same consideration as above applies, and a
description set exposed by repository B to express a relationship with a resource
described in a description set exposed by repository A, and vice versa, the additional
factor to consider here is the compatibility of the domain models underpinning the
two DCAPs.
If for example, repository B deploys a FRBR based model, the expectation may be
that the type of relationship in question is expressed between two FRBR Works or
between two FRBR Expressions. This is the case, for example, with whole-part
relationships in FRBR. But if the data exposed by repository A is based on a different
model which does not include concepts of Work and Expression, then it may not be
clear how those relationships can be expressed using the FRBR-based model.. The
owner of repository B is faced with the choice of remodelling/redescribing the
resources within the FRBR model, or trying to adapt their model to encompass some
more general relationship types.
3
For examples of this sort of scenario, the Time-Based Media project has a use case
involving relationships between still images and videos from which the stills are
taken; and similarly, although the Scholarly Works Profile addressed relationships
between paper and presentations (as distinct Works), it may also be useful to capture
the relationships with audio and video captures of presentations.
4
And again in this scenario, the aggregator’s knowledge of the vocabularies and
structural patterns specified by the two DCAPs means that query patterns can be
predictably constructed.
What is different in this case, however, is that, depending on the differences between
the two profiles, the aggregator may need to apply multiple query patterns across the
merged data.
If the individual DCAPs are based on a common “domain model”, and use some
common vocabulary of metadata terms, even if the profiles also differ in some more
specific aspects. then some common query patterns may apply alongside some
“profile-specific” patterns. The greater the divergence between the two domain
models, the less likely it is that common query patterns will be usable, and the more
likely it is that “profile-specific” patterns are required.
As the number of DCAPs and the number of different models increases, so the
number of different query patterns to be managed by the aggregator increases. This
doesn’t make such querying impossible: it just increases the level of complexity to be
managed by the aggregator.
5
In these contexts, the similarities and differences between the models underpinning
the different DCAPs may become significant. This might be addressed in various
ways:
• Is it possible to simply extend the individual models?
• Is it necessary/desirable to try to “harmonise” those individual models?
• Is it necessary/desirable to map from those individual models into a separate
model which does try to address the full range of resource types within a single
model? If so, what are the options? Simple Dublin Core? BIBO? FRBR? CIDOC
CRM? Something else?
6
1
Scholarly Works Application Profile (SWAP)
http://www.ukoln.ac.uk/repositories/digirep/index/Eprints_Application_Profile
2
Images Application Profile (IAP) http://www.ukoln.ac.uk/repositories/digirep/index/Images_Application_Profile
3
Geospatial Application Profile (GAP) http://www.ukoln.ac.uk/repositories/digirep/index/Geospatial_Application_Profile
4
DCMI Abstract Model (DCAM) http://dublincore.org/documents/2007/06/04/abstract-model/
5
RDF Semantics http://www.w3.org/TR/2004/REC-rdf-mt-20040210/
6
The Singapore Framework for Dublin Core Application Profiles http://dublincore.org/documents/2008/01/14/singapore-
framework/