|
Introduction:
The emergence of global networking has made it
possible to use the Internet as a global file system managed with a
globally distributed ‘operating system’ – the Web. The foundation protocols of the
Internet include the Domain Name System, perhaps the world’s most
ubiquitous and successful registry system. As Internet and Web protocols have
evolved, the need for registries for a variety of purposes has become
evident, including the management of metadata terms. The Dublin Core
Metadata Registry is one approach to meeting such needs. Emerging from a common interest in
the Dublin Core Metadata Initiative (DCMI) community, the DC-Registry
working group [2] has, since December 1999, focused on identifying the
functional characteristics for a metadata registry that meets the needs
of our community, and providing a forum for open discussion of related
issues.
These discussions have benefited from the
perspectives of many researchers and practitioners in the DCMI
community. The development of
the software itself has been largely concentrated in OCLC Research, with
significant input from UKOLN (the UK Office for Network Learning) at the University of Bath, and the Library and
Information Science Department of the University
of Tsukuba
. As an Open Software initiative,
OCLC makes the software available under the Dublin Core Public License
Version 1.0 [3].
The Dublin Core Metadata Registry is designed and
deployed as a multilingual registry of metadata terms. It is not configured as a schema
registry, though as a general purpose application designed around
Resource Description Framework (RDF) [4] entities, it could be used to
manage RDF schemas. Other
projects, for example IEMSR at UKOLN (Johnston,
2005, [5]) take this approach, and there are commercial applications
designed for schema management as well (SchemaLogics, for example [6]).
The registry is intended to serve as a discovery
mechanism and resolution service, with the goal of promoting the reuse of
existing terminologies represented in multiple languages. The registry is packaged
with the Dublin Core metadata vocabulary, but is not limited to this, or
any given vocabulary.
Additional metadata terms can be included at the discretion of the
implementer. The only
limitation regarding what metadata can be imported into the registry is
that it must be represented within the registry in RDF because of the
reliance on RDF for internal registry data encoding.
2. Architecture of
the DCMI Registry
The
DCMI metadata registry is a distributed application in the sense that it
is a collection of registries loosely federated with other metadata
registries running at other locations. Each of the DCMI registries shares
a common data format, user interface, and inter-registry communication
features that enable them to cooperate and share metadata declarations in
a loose federation.
The
DCMI metadata registries are being deployed to serve the needs of
communities of practice that differ according to domain, policy
constraints, or natural language.
The current DCMI registry design dates back to the publication of
“Plans for a distributed registry of Dublin
Core in multiple languages” (Baker, 1998). The motivation for distributing
the registry is simply to support the needs of communities of practice in
tailoring
their metadata implementations to meet local needs and constraints while
sustaining high degrees of compliance with an international
standard.
Satisfying this need is a principal goal of the DCMI metadata
registry.
DCMI
metadata terms are intended to be broadly applicable across domains. Extensibility was among the
original design criteria, supporting the need to define additional, more
specific terms when necessary.
The design of the registry reflects this extensibility
requirement. The Dublin
Core community classifies terms as elements, element-refinements,
encoding schemes and controlled vocabulary terms. Dublin
Core practitioners often conceptualize DCMI terms in these logical
groupings. The registry
supports this by providing pre-defined views of the data that match these
conceptualizations.
DCMI
registry implementations may differ from one another in a number of ways,
depending on the communities they serve, the languages they support, and
the number of local extensions they implement. However, from architectural and
organizational perspectives they share important similarities. First, each registry runs the same
base software and supports the same interfaces to its content.
Additionally, each of the
communities
these registries serve is comprised of:
• Read-only users.
These include both the humans and applications that are the
primary consumers of the registry content. Additionally, the read-only users
provide feedback to the registration authority regarding change requests
to the registry content.
• A registration authority, responsible for approving
registry content. For
example, the DCMI Usage Board is the registration authority for the
registry available at the Dublin Core Web site [1]. The Usage Board evaluates proposed
new terms that are suggested by the larger Dublin Core community. Approved terms are then passed to
the registry steward for inclusion in the metadata registry.
• Stewards are responsible for application support and
maintenance. Their role is
limited to the development, support and ongoing maintenance of the
registry software. They rely
on the registration authority for decisions regarding the actual registry
content.
From a technical perspective, the registry is a
server-side Java application.
As such, it can run on any platform that supports Java. It is built entirely on open-standards
and is distributed as open-source.
A number of databases are supported, including PostgreSQL, MySQL,
and Oracle [7].
The registry relies on open-source distributions and
public standards for managing registry contents, including:
• Extensible Markup Language
(XML) for parsing, serializing and exchanging registry content. All registry output is in XML
format. Xerces [8] is used
for processing XML.
• RDF and Resource Description
Framework Schema (RDFS) [9] for defining metadata terms and their
associated properties. Registry
content is represented internally as RDF statements. Jena
[10], an open source Java framework for Semantic Web applications, is
used for reading, writing and parsing RDF and ontology languages.
• Extensible Stylesheet Language
Transformations (XSLT) [11] for presenting search and browse results to
users in a human-friendly (Web browser) format.
• The design of the registry is
strongly oriented towards Web services, providing application access to
registry content via both Simple Object Access Protocol (SOAP) and
Representational State Transfer (REST) [12]. Axis, an Apache open-source
project for creating SOAP-style Web services, is used for processing the
SOAP-style application interface[13]. The REST-style application
interface simply relies on HTTP Post and Get.
Figure 2: DCMI registry software
components
Usability
assessment has been limited to discussions within the Registry Working
Group.
These
discussions were conducted on the (publicly available) working group
mailing list and focused primarily on functional requirements and
technical solutions. Various
prototypes were developed, based on these discussions, to facilitate this
effort (Heery and Wagner, 2002).
The registry currently supports 25 languages,
including translations of both the user interface labels and the Dublin
Core terms. The translations
of Dublin Core terms have been done largely on a volunteer basis. These translations range in
authority from those sponsored by national libraries or other national
information agencies to the volunteer efforts of single individuals [14]. Much of this work has been
contributed by active participants in the DC-Registry working group and
the DC-Localization and Internationalization Working Group [15]. The translations, being voluntary,
vary with regard to completeness and accuracy, and continue to evolve as
new translations become available.
Provenance information regarding the source of term definitions is
provided within the registry for all terms. The English language rendition
comes from the formal standard (ISO 15836:2003) [16].
The DCMI Abstract Model is “a reference model
against which particular DC encoding guidelines can be compared”
(Powell, et al., 2005).
The Abstract Model was advanced in 2003, and became a formal DCMI
Recommendation in March of 2005.
The Abstract Model articulates the architecture of DC metadata and
the DCMI Registry is intended to be the authoritative source of
information about DC metadata terms.
As such, these two entities must be aligned. While design and implementation of
the DCMI Registry predates the abstract model, no conflicts between them
have been identified.
3. Functional
Characteristics and Benefits of the DCMI Registry
The functionality of metadata registries is a subject
of active research and can be expected to evolve over time. At present, registries are
primarily resolution services, resolving identifiers to information about
resources (either physical or conceptual). In the case of metadata
registries, term identifiers are resolved to information about
terms. It is useful to
compare the benefits of such a system to more direct means of declaring element sets, such as schemas and static
pages (either electronic or paper).
Schemas are static representations of terms, their
properties, and their relations to one another. They are meant for
machine interpretation, and are commonly written in either XML or RDF
schema languages. They
require an understanding of modeling techniques as well as experience
with schema languages to maintain and interpret.
Static pages (HTML or PDF for example) require no
specialized infrastructure or systems development. They are simpler to update and
display and are easier to use for many simple look-up tasks.
Metadata registries must add value that substantially
exceeds that provided by static representations (schemas or static pages)
in order to be widely deployed and used. Metadata registries provide
people and applications with an authoritative source of in-depth
information about terms and their relationships to other terms within a
given set of vocabularies.
They are intended to serve the needs of more than one class of
users,
including:
• organizations that maintain
large, and/or multilingual metadata vocabularies,
• metadata applications,
communicating with the registry via Web services,
• schema and ontology designers,
and
• creators of metadata instance
data.
Metadata registries meet these
needs in a number of ways:
Discovery: Discovery is a
prerequisite to reuse.
Designers of metadata applications need
tools that enable them to easily discover terms that
are already in use and that can simply be adopted. Besides reducing costs, high
levels of re-use are an important ingredient for standardization and
interoperability. While a
static web page or schema can provide this information, the user must
first find the document, and then determine whether or not it is the
latest version (particularly difficult if the document is
paper-based). Metadata
registries, deployed within recognized communities of practice, provide
known sources for up-to-date information about terms. This information is available in
formats suitable for humans and applications, and can be provided in a
number of natural languages.
Authority:
Registries can be managed according to formal policies determined by
responsible organizations, including access
constraints and formal measures of term currency or status. The existence and management of
such policies support two key attributes necessary for the success of the
emantic Web; trust and provenance.
The “Web of Trust” (Berner-Lee, 1998) recognizes that
authority of information is an increasingly important issue. Metadata registries provide a
useful tool for assuring unambiguous, up-to-date information about
metadata terms, their provenance, and available translations.
Dynamic representation of
information: Registries can
offer customized representations
of a given metadata set to satisfy different classes
of users:
• Multiple languages – A
registry can provide term definitions in any number of supported natural
languages. This is
particularly important in environments where metadata designers,
creators, and users need to manage metadata in multiple languages.
• Multiple encoding – The
expression of metadata is bound to a chosen encoding, and while in an
ideal world there would be only one, the reality is that there are
multiple encoding formats, including, HTML (and variants), XML, RDF and
Notation 3 (N3) (Berners-Lee, 2001). The DCMI registry provides
quick-access links to a variety of encoding formats for terms, including
usage examples.
• Customized conceptual views
– Organizing metadata terms into collections serves two useful
purposes; it facilitates term management, and enables communities to view
their data in ways that are more meaningful to them.
Metadata
registries include information about definitions and syntax of metadata
terms but they must also be flexible enough to provide logical views of
that information that promote discovery and comprehension.
Information suitable for both
humans & machines: Documents formatted for humans,
while being easier for people to use in some cases,
cannot be easily processed by applications. The ability of a registry to
provide applications with information about terms via Web services opens
up opportunities to increase the effectiveness of metadata systems
through automation.
For example, consider a hypothetical application that
harvests metadata related to ducational materials, and processes this
information based on intended audience. An application such as this might
be aware of specific terms, such as ‘audience’, and use these
terms to aggregate instance data based on the value of this term. What happens though when
‘audience’ is further refined, as happened when the
element-refinements ‘educationLevel’ and ‘mediator’ were
later added. Such an
application would very likely also be interested in these new terms. How would it discover they exist,
and that they were related to ‘audience’. Several alternatives are possible:
• An application can be
hard-coded, or parameter-driven, for the terms it uses. Such applications would require
someone to discover the new terms, recognize their relationship to terms
currently being used, and then modify the application to use the new
terms. This scenario is prone
to failure, and falls short of our expectations for the Semantic Web.
• An application can be
schema-aware, and capable of parsing schemas that are likely candidates
for new terms. However, this
level of application sophistication requires additional resources and
technical expertise to develop and support. Second, and more importantly,
schemas are commonly versioned.
It is fairly common practice to add new terms to a new version of
a schema, rather than changing an existing schema. This is intended to reduce the
potential for breaking existing applications, but also leaves
applications unaware of new term additions.
• An application could use a Web
services interface such as are provided by the DCMI registry to identify
new terms and their relationship to other terms. This is more than sufficient for
the hypothetical application used in this example to recognize that new terms
were added, and to understand their relationship to existing terms. Such ‘registry-aware’
applications won’t require code changes or human intervention to
evolve with the terminology they use. One example is the DCMI
translation tool [17], which uses the registry’s application
interface to collect and present information about the natural language
properties (label, definition and description) associated with each DCMI
term.
Illustration of complex formal
relationships among terms:
Most systems embody
complicated metadata where there are relationships
among terms. This is true for
any DC metadata that has qualifiers (schemes for the constraint of term
values, or element refinements that narrow the meaning or scope of a
term). As the influence of
ontology research increases, these relationships will become more complex
and will require tools that enable practitioners to easily visualize the
nature of these relationships.
Registries are better suited to this task than static pages. They can display data in ways that
promote ease of comprehension.
For example, consider the relatively simple relationships enabled
with RDF Schema (RDFS), which includes a ‘subPropertyOf’
assertion. Using RDFS,
metadata practitioners can make explicit that ‘abstract’ and
‘tableOfContents’ are refinements of
‘description’.
However, the opposite is not true. RDFS does not provide terminology
that enables practitioners to explicitly assert that
‘description’ is further refined by ‘abstract’
and ‘tableOfContents’.
This must be inferred.
Such relationships can be made more explicit in registries. This
will become a prominent benefit of registries as ontology languages (such
as the Ontology Web Language, or OWL [18]) and reliance on logical
inference using ontologies becomes more widespread.
4. Relationship of
the DCMI Registry to ISO/IEC 11179
ISO/IEC
11179 [19] is a standard that describes a model for metadata
registries. It is comprised
of multiple arts, with varying levels of conformance. These include framework,
classification, metamodel (a basic set of properties for describing
registry content), data definition guidelines, naming and identification
of data elements and content registration. The goal of these specifications
is to provide a common framework for registries that will insure
increased likelihood that:
• Registered items can be uniquely identified.
• Properties for describing registry content are
unambiguous.
• Registry content can be unambiguously mapped between
the metamodel and the implementation.
The
overriding goals of the 11179 specification are to promote the
understanding and sharing of metadata erms and definitions. This is also true for the DCMI
metadata registry. However,
while the DCMI registry shares a similar mission with the 11179 standard,
it differs somewhat in its approach, specifically with regards to
technology.
Typically,
11179-compliant implementations have been XML-centric in their
approach. XML is pervasive,
and widely successful as a data exchange encoding. However, XML does not lend itself
well to two key aspects of 11179 compliance:
1)
content properties that are unambiguous, and
2)
insuring the uniqueness of identifiers. Part 3 (Registry Metamodel and
Basic
Attributes),
part 4 (Formulation of Data Definitions) and part 5 (Naming and
Identification Principles) address these issues and provide guidance for
achieving compliance using an XML-centric approach.
The
DCMI registry also relies on XML as a means of data exchange, but uses
additional technologies (RDF and XML Namespaces) that are layered on top
of XML. The issue of insuring
unique identifiers for each registered item is addressed by using fully
qualified term names for identifiers. Term names, such as:
http://purl.org/dc/elements/1.1/creator
consist
of a namespace:
http://purl.org/dc/elements/1.1/
and a
local term name:
creator
Properties
described using RDF have three key advantages over simple XML tags:
1) they
take advantage of XML namespaces,
2) they
have attributes which describe their essential nature, and
3)
these attributes can be processed by applications.
Together, these features insure that these properties
are unambiguous and that their semantic meaning is clearer. For example, consider a
geographic area described with the extent property. This property is identified
with a URL that resolves to a series of attributes describing it and its
relationship to other properties:
http://purl.org/dc/terms/extent
RDF-aware applications can process this information
and identify a number of assertions about this property. The attributes describing the term
in the schema are publicly declared assertions. Specifically:
The English rendition of the label is: Extent
The English rendition of the description is: The size or duration of the
resource.
The source for this term definition is: http://purl.org/dc/terms/
It has an RDF Type of: Property
It has a dc term type of: element-refinement
It has an associated version: extent-002
It has a date issued of: 2000-07-11
It was last modified on: 2002-06-15
It is related to and refines: format
However, without the explicit declaration of
attributes as is provided with RDF, applications must rely on a
pre-arranged understanding of, or assumptions about, the meaning and
nature of simple XML tags, such as ‘<extent> or
<geographicExtent>’.
The ISO 11179 community is currently exploring ways
to expand the specification to include greater utilization of XML
namespaces and to provide additional support for the semantic
understanding of registered items and their properties. Part of this effort is being
undertaken by the Extended Metadata Registry Project (XMDR) [20]. One of the key issues being
considered is how to incorporate extensions to the specification to
include support for technologies, such as RDF and OWL, which promote
unambiguous semantic specification.
This increased focus on declarations of semantics is reflected in
the 2005 Open Forum on Metadata Registries conference [21], an annual
international conference, the
theme of which, was “Semantic Interoperability: Where
Meaning Meets
Metadata”
The DC-Registry working group recognizes the
importance of a standard approach for metadata registries. They have closely followed the
work being done on the ISO/IEC 11179 specification, and where possible,
have sought opportunities for mutual cooperation. It is expected that the extensions
being considered by the XMDR project will help bring the two communities
closer together.
5. Related Registry
Activities
In an era of increasing importance of
machine-to-machine transactions, registries will be important for
exchanging well-structured information and schemas. Kotok (2003) describes several
examples, including Universal Description, Discovery and Integration
(UDDI) and electronic business using eXtensible Markup Language (ebXML),
registries that promote discovery of Web services and product information
[22].
Metadata registries, in turn, provide resolution
services for terms or abstract concepts, resolving identifiers to
information about metadata terms used to describe products and services
or intellectual assets. Their
goal is not to provide access to products or services, but rather to
promote understanding of metadata terms used to describe products and
services. The DCMI Metadata
Registry falls into this category.
Application profiles are an important part of the
changing landscape of metadata standards, affording a necessary mechanism
for combining terms from several schemas into a composite schema that is
tailored to local functional requirements (though an application profile
may be comprised of terms from a single schema as well). An application profile is
generally rendered in the form of a compound schema (as described in
Heery and Patel, 2000).
Operationally, application profiles require both documentation and
standards of community practice to become useful and effective. The development of application
profiles and their ancillary support is still as much a topic of research
as practice, but as they proliferate, they will be prominent entities in
metadata
registries.
The variety of metadata registries is substantial,
reflecting the research interests and operational needs of the many
organizations that have undertaken them. They differ according to the
underlying technology strategies, desired functionality, and
accessibility by humans, applications, or both. Prominent examples of metadata
registry applications span a variety of domains.
The Environmental Data Registry provides information
used to describe environmental data used by local, state and U.S.
government agencies. It is
intended to serve as an authoritative and comprehensive source of
information about terminology used by these agencies. This registry is part of a larger
System of Registries (SOR) implemented by the U.S. Environmental
Protection Agency [23].
The U.S. Department of Defense Metadata Registry and
Clearinghouse [24] is intended to provide a common source of information
about metadata terms and related technologies that are used within the
defense industry. The
registry is designed to promote interoperability and reuse of metadata
and related software within authorized agencies. Details of the contents of this
registry are no longer accessible to the general public.
The METeOR registry serves the Australian Institute
of Health and Welfare (as described by Braddock, 2005) This registry is a
replacement for the Knowledgebase Registry Like its predecessor, it
registers metadata related to Australian social services, specifically:
health, community services and housing assistance. It includes browse and search
interfaces and is based on the ISO/IEC 11179 specification [25].
The European Library (TEL) Registry [26] is a
collection of terms and properties designated for use in TEL application
profiles (van Veen and Oldroyd, 2004). TEL is a cooperative effort
involving 8 European national libraries and the Italian Central
Cataloging Institution (ICCU).
Their goal is to provide an advanced resource discovery service
for researchers that goes beyond enumeration of available terms, and
includes information about application profiles, schemas, and support for
data entry forms and proposing of new elements.
The CETIS registry, developed and maintained by the
Centre for Educational Technology Interoperability Standards, is intended
to “help people in the UK HE (higher education) & FE
(further education) community to find out about terms used in the field of
learning technology standards” [27]. This registry serves as a
reference tool, providing browse and search access to a variety of
definitions related to technology.
Definitions are provided collaboratively by members of the CETIS
special interest groups (SIGs).
The German Metadata Registry provides “an
overview of the metadata efforts and implementations within Germany
and German-language areas” [28]. This registry includes term
definitions, documentation, links and other material related to a number
of subject specific domains, including education, medicine, physics and
mathematics. Both German and English languages are supported.
The Development of a European Service for Information
on Research and Education (DESIRE) Metadata Registry [29] was among the
first registry metadata projects.
This registry was developed by UKOLN and is intended to serve as a
discovery and navigation tool for a variety of metadata resources,
including namespaces, registration authorities, application profiles and
cross-vocabulary mappings between terms. It relies on XML as
the underlying technology and is the basis for a
number of evolving registry activities described below. While each of these projects
continue to have a web presence and a functional registry, not all are
currently active.
• The SCHEMAS Registry [30]
includes terms and metadata activity reports related to projects funded
by the IST Programme and other European national initiatives. It is implemented as a
human-readable registry designed primarily to inform metadata designers.
• The CORES registry (Heery, et
al., 2003), is a registry of metadata vocabularies that focuses on the
sharing and reuse of metadata terms, and the creation of application
profiles [31]. It builds on
work done in the SCHEMAS project.
Among the practical outcomes of this activity is a resolution
among several metadata organizations to identify their metadata terms
with Uniform Resource Identifiers (Baker & Dekkers, 2003).
• The Metadata for Education
Group (MEG) Registry [32] is both an extension, and a rewrite, of the
DESIRE registry. Unlike
DESIRE, the MEG registry is RDF based. It is intended to serve the needs
of the MEG group by providing a known source of information related to
education-specific semantics, application profiles and metadata
specifications. The registry
also includes a Java client designed to facilitate application profile
creation (Heery et al., 2002).
• The Information Environment
Metadata Schema Registry (IEMSR) [33] is a recent registry project that
builds on the MEG registry.
It is expected to include both the Dublin Core and IEEE LOM [23]
metadata vocabularies and serve as a known source of information for
education-related application profiles (Johnston,
2005).
6. DCMI Registry Deployment Experience At
present
five DCMI metadata registries have been deployed, at
the following locations:
• OCLC Online Computer Library Center, Dublin, Ohio, USA
• The University of Tsukuba, Japan
• The University of Goettingen, Germany
• The Library of the Chinese Academy of Science, (Beijing) China
• The National Library of New Zealand, New Zealand
The registry deployed at the Library of the Chinese Academy of Science in Beijing
is a recent addition. It is
expected to play a role in a project to link all of the Chinese digital
libraries. Each of the remaining registries is currently being used
primarily for research.
The registry application statistics offer a valuable
source of insight regarding registry functionality, and its
usefulness. Table 1 provides
a summary of pages served by the Web (human) interface of the Dublin, Ohio
registry. These statistics
cover a period of activity spanning November 1, 2004 through April 30,
2005. A total of 137,581
pages were accessed during this period. Registry users do not identify themselves and hence
cannot be questioned about their reasons for using the registry, however
logs reveal certain usage patterns.
For instance:
• The registry provides two Web
interfaces, one for browsing the registry content, and one for text
searching. The browse feature
was used over 10,000 times, compared with only 272 search requests,
indicating a clear preference for browsing over searching.
• There were 31,930 requests for
term-level detail. This includes term-level information in all of the
supported languages.
• There were 47,613 accesses for
alternate encoding views (term information encoded in RDF, N3, or
N-triple). These data
encodings are accessible as links from the term detail views formatted
for human reading, but can also be accessed directly via URL. The large number of page views for
these encodings is a good indication of the interest that registry users
have in Semantic Web technologies.
• There were 1,262 requests to
set or change language preferences.
The default language for this registry is English. The large number of calls to
change language preferences is indicative of a significant number of
international users whose first language choice is not English.
• The term detail view provides
quick links to usage examples (7,669). The large number of access
requests indicates the importance of this feature.
Request Type
Page CountPercentage
Browse registry content
10,415 7.5%
Search registry content
272 0.1%
View item detail
31,930 23.2%
View alternate encoding (RDF, N3, etc.) 47,613 34.6%
Set or change language preferences 1,262 0.9%
View usage examples for a term
7,669 5.5%
Canonical view of term
4,444 3.2%
Other (i.e., Provenance information) 33,976 24.6%
Table 1: Page summary
statistics: DCMI Registry
2004.11.01 – 2005.04.30
Registries are expected to serve applications as well
as humans. The DCMI Registry
supports two application interfaces, both based on Web services. One is SOAP-based and the other is
REST-based. The registry
statistics provide insight into these two interfaces and how they are
being used. The overwhelming
majority of application use comes from the REST-style services. During the period from November 1,
2004 through April 30, 2005 there were 4,841 calls to the REST services,
versus 1,727 SOAP-style calls, possibly reflecting the greater ease of
implementation of REST services.
The application interfaces provided are designed to
satisfy a number of different information request types. This was considered an important
functional requirement, and according to an informal survey recently
conducted within the DC-Registry Working Group, is still perceived as one
of the most important features of metadata registries. Based on this information one
might expect the usage statistics to show patterns indicative of use by
applications. The statistics show interest in the application interfaces,
but the use-patterns currently reflect experimentation, rather than
actual use.
7. Barriers to
Installation and Adoption
Organizations adopt technology because it promises to
add value: to improve delivery of goods and services or reduce the cost
of providing them. Identifying this value at the present time is a
largely qualitative assessment.
Partly this is due to the fact that the metadata milieu is
evolving. The value of
metadata registries will become more evident as the difficulties
associated with large deployments of heterogeneous metadata systems
emerge. There is growing
evidence of the problems of uncoordinated metadata assignment and design
in systems which attempt to integrate metadata from many sources. For example, studies of the
NSF-funded National Science Digital Library program have identified
inconsistent metadata as a major impediment to effectiveness (Dushay and
Hillmann, 2003). Metadata
registries can help address such deficiencies.
Organizational commitment is another important factor
that has to be considered.
Deploying and sustaining software systems entails substantial
costs; organizations must make a significant and persistent commitment to
deploy them. There must therefore be a strong business case to support
their adoption.
Deploying a metadata registry requires a thorough
understanding of a variety of factors:
• What is the scope of the
registry .
• Who is it intended to serve.
• Is there an organization to
establish and administer registry policies.
• Are there technical resources
available to manage the day-to-day technical aspects of running the
application.
• How should the data be
organized to best promote its comprehension. Metadata should be organized so
that it can be viewed in ways meaningful to the registry users.
The degree of technical expertise required to
install, customize, and maintain a registry, and the amount of planning
required to make it productive, can be substantial.
8. Prospects for
Further Development
Development of metadata registry infrastructure
remains at an early stage, meeting primarily administrative needs, and
with human users as the primary consumer. Today, metadata registries are
essentially a resolution service, resolving term identifiers to
information about terms. However,
one can imagine a metadata environment that benefits from a wider variety
of services and a greater degree of application-to-application
communication. Possible
scenarios include:
• Automated crosswalks and
mappings. The Dublin Core
metadata registry is already capable of registering metadata terms that
can be encoded in RDF, including terms from other metadata
standards. However, simply
including terms from more than one metadata standard within the same
registry does not solve the interoperability problems that exist between
standards. A means to
automate the mapping of terms between standards is needed and is a good
candidate for future development.
• Querying and use of
‘arbitrary’ schemas (application profiles) by
applications. Application
profiles, and how to generate them in a machine-readable format, is an
ongoing research activity.
When a standard for generating machine-readable application
profiles emerges and is adopted it is conceivable that the Dublin Core
registry will include support for application profiles.
• Automated invocation of
authority management, harvesting opportunities, and metadata
triggers. The cost of
creating instance metadata is a large impediment to the effectiveness of
metadata systems and resource discovery and management.
Efforts to reduce that cost must rely largely on
industrializing the creation of metadata through more effective tools and
automation. As such
techniques emerge, they will benefit from gathering authority-controlled
data from a variety of sources.
This will only become possible when applications can discover both
the content and structure of such data without human intervention, and
metadata registries will likely be an important part of this infrastructure.
• Automated harvesting of metadata
terms. Registering terms in
the current DCMI registry involves importing terms using one of the
import tools provided with the software. Terms must be identified and then
selected for importing. In
the future it will be desirable to automate harvesting of metadata
repositories, extracting and automatically registering relevant terms.
9. Conclusions
The DCMI Registry has evolved over a period of years
to meet the needs of the DCMI community, which have ranged from
authoritative management of the DCMI vocabulary, to public browse and
search capability, to application interfaces. Feedback and participation from
the community have helped steer the design and implementation, and the
architecture and functionality have evolved to a model very much in
keeping with the distributed idiom that currently dominates Internet
developments. Distribution of
management and function incurs some management costs, but confers
technological resilience as well as meeting important policy requirements
that are essential components of global information systems.
Nonetheless, metadata registries have not yet become
an integral part of the metadata infrastructure of the Web. We expect this integration will
increase as multi-lingual or cross-domain metadata applications are
deployed on a large scale, and the benefits for automated management
support becomes more evident.
References
Baker, Thomas (1998). “Plans for a distributed
registry of Dublin Core in multiple languages.” Dublin
Core Working Draft, 1998-10-28.
http://dublincore.org/documents/1998/10/28/distributed-registry/.
Baker, Thomas
and Dekkers, Makx
(2003),
“Identifying Metadata Elements
with URIs: The CORES Resolution”
D-Lib Magazine, Volume 9 Number 7/8.
http://www.dlib.org/dlib/july03/baker/07baker.html.
Baker, Thomas; Dekkers, Makx; Heery, Rachel; Patel,
Manjula, and Salokhe, Gauri
(2001)
“What Terms Does Your Metadata Use. Application Profiles as
Machine-
Understandable Narratives”. Journal of Digital
Information, Volume 2 Issue 2
Article No. 65, 2001-11-06
http://jodi.ecs.soton.ac.uk/Articles/v02/i02/Baker/
Berners-Lee, Tim (2001). “Notation 3: An RDF
language for the Semantic Web”
2001-
11-27.
http://www.w3.org/DesignIssues/Notation3.html
Berners-Lee, Tim (1998). “Realising the Full
Potential of the Web.”
W3C Website.
http://www.w3.org/1998/02/Potential.html.
Braddock, David (2005). “METeOR: a metadata
registry implementation experience”
Open Forum 2005 on Metadata
Registris.
http://www.berlinopenforum.de/download/Braddock_David.zip
Dushay, Naomi. & Hillmann, Diane (2003).
“Analyzing Metadata for Effective Use and
Re-Use.”
DC-2003 Dublin
Core Conference: Supporting Communities of Discourse and
Practice – Metadata Research
and Applications. September 28 -October 3, 2003.
http://www.siderean.com/dc2003/501_Paper24.pdf
Heery, Rachel, Johnston,
Pete, Beckett, Dave, and Steer, Damien (2002). “The MEG
Registry and SCART: Complementary Tools for Creation,
Discovery and Re-use of
Metadata Schemas” DC-2002: Metadata for
e-Communities: Supporting Diversity and
Convergence. Florence, Italy.
October, 13-17, 2002.
http://www.bncf.net/dc2002/program/ft/paper14.pdf
Heery, Rachel; Johnston, Pete; Fülöp, Csaba
and Micsik,Andras (2003). “Metadata
schema registries in the partially Semantic Web : The
CORES experience.” DC-2003
Dublin Core Conference: Supporting
Communities of Discourse and Practice – Metadata
Research and Applications.
September 28 -October 3, 2003.
http://www.siderean.com/dc2003/102_Paper29.pdf
Heery, Rachel and Patel, Manjula (2000).
"Application profiles: mixing and matching
metadata schemas" Ariadne, Issue 25,
24-Sep-2000.
http://www.ariadne.ac.uk/issue25/app-profiles/intro.html.
Heery, Rachel and Wagner, Harry (2002). “A
Metadata Registry for the Semantic Web”
DLib Magazine. Volume 8, #5 May, 2002.
http://www.dlib.org/dlib/may02/wagner/05wagner.html.
.
Johnston, P. (2005). "What Are Your
Terms." Ariadne, Issue
43, April 2005.
http://www.ariadne.ac.uk/issue43/johnston/.
Kotok, Alan (2003). “'Metadata Rules' - a
report from the Open Forum on Metadata
Registries.” WebServices.org 2003-02-24.
http://www.webservices.org/index.php/ws/content/view/full/2873.
Powell, Andy, Mikael Nilsson, Ambjörn Naeve,
Pete Johnston (2005). “DCMI Abstract
Model.” DCMI Recommendation. 2005-03-07.
http://dublincore.org/documents/abstract-model/.
van Veen, Theo and Oldroyd, Bill (2004).
“Search and Retrieval in The
European Library: A New Approach”
D-Lib Magazine, Volume 10 Number
2.
http://www.dlib.org/dlib/february04/vanveen/02vanveen.html.
NOTES
[1] The Dublin Core Metadata Registry is available
at:
http://dublincore.org/dcregistry/.
[2] The DC-Registry working group page is available
at:
http://dublincore.org/groups/registry/.
[3] DCMI public license is available at:
http://dublincore.org/dcpl/.
[4] The Resource Description Framework (RDF) is a
family of W3C specifications
defining encoding standards to support Semantic Web
applications.
http://www.w3.org/RDF/.
[5] The JISC IEMSR project is available at:
http://www.ukoln.ac.uk/projects/iemsr/.
[6] SchemaLogic Corporate homepage:
http://www.schemalogic.com/page.
[7] PostgreSQL is available at: http://www.postgresql.org/.
MySQL is available at: http://www.mysql.com/.
Oracle is
available at: http://www.oracle.com/index.html.
[8] Xerces is available at:
http://xml.apache.org/xerces2-j/.
[9] RDF Vocabulary Description Language 1.0: RDF
Schema. W3C Recommendation 10
February 2004.
http://www.w3.org/TR/rdf-schema/.
[10] Jena
is an open source project originating in the HP Labs Semantic Web
Programme, available at: http://jena.sourceforge.net/.
[11] XSL Transformations (XSLT). Version 1.0. W3C
Recommendation 16 November
1999. http://www.w3.org/TR/xslt.
[12] SOAP Version 1.2. W3C Recommendation 24 June
2003.
http://www.w3.org/TR/soap/.
The REST protocol is described in chapter 5 of the
doctoral dissertation by Roy Fielding:
“Representational State Transfer
(Rest)”, in Architectural Styles and the design of
Network-based
SoftwareArchitectures. Univeristy of California, Irvine
, 2000.
http://www1.ics.uci.edu/%7Efielding/pubs/dissertation/rest_arch_style.htm.
[13] Axis is an implementation of the SOAP protocol,
available at:
http://ws.apache.org/axis/index.html.
[14] Sources and acknowledgements for Dublin Core
registry term translations are
available at:
http://dublincore.org/dcregistry/pageDisplayServlet.page=help_en-US.xsl#H7.
[15] Dublin
Core Localization and Internationalization Working Group.
http://dublincore.org/groups/languages/.
[16] Information and documentation - The Dublin Core
metadata element set (ISO
15836:2003
http://www.iso.org/iso/en/CatalogueDetailPage.CatalogueDetail.CSNUMBER=37629&s
copelist=PROGRAMME.
[17] The
DCMI translation tool is available at:
http://wip.dublincore.org/translate/.
[18] OWL Web Ontology Language. W3C Recommendation. 2004-02-10
http://www.w3.org/TR/2004/REC-owl-features-20040210/.
[19] ISO/IEC 11179 standard for metadata
registries.
http://metadata-standards.org/11179/.
[20] Extended Metadata Registry Project (2005-05-19)
is available at: http://xmdr.org/.
[21] Eighth International Open Forum on Metadata
Registries. April 6-8, 2005.
http://www.berlinopenforum.de/.
[22] Universal Description, Discovery and Integration
(UDDI) is available at:
http://www.uddi.org/.
Electronic Business using eXtensible Markup Language
(ebXML) is available at:
http://www.service-architecture.com/web-services/articles/ebxml_registry.html.
[23] Environmental Data Registry is available at
http://www.epa.gov/edr/ and
the
System of Registries (SOR) is available at:
http://www.epa.gov/sor/.
[24] The U.S. Department of Defense Metadata Registry
and Clearinghouse is
available
at:
http://diides.ncr.disa.mil/mdregHomePage/mdregHome.portal.
[25] The
METeOR registry is available at:
http://meteor.aihw.gov.au/content/index.phtml/itemId/181414.
The Knowledgebase Registry is available at:
http://www.aihw.gov.au/knowledgebase/index.html.
[26] The European Library (TEL) Registry is available
at:
http://krait.kb.nl/coop/tel/handbook/tel_reg_v1.3.html.
[27] The CETIS registry is available at.
http://www.cetis.ac.uk/encyclopedia.
[28] German Metadata Registry Project is available
at:
http://www.mpib-berlin.mpg.de/dok/metadata/gmr/gmr1e.htm.
[29] Development of a European Service for
Information on Research and Education
(DESIRE) Metadata Registry is available at:
http://desire.ukoln.ac.uk/registry/index.php3.
[30] The SCHEMAS Registry is available at:
http://www.schemas-forum.org/registry/.
[31] The CORES registry is available at:
http://www.cores-eu.net/registry/.
[32] The Metadata for Education Group (MEG) Registry
is available at:.
http://www.ukoln.ac.uk/metadata/education/regproj/.
[33] The Information Environment Metadata Schema
Registry (IEMSR) is available at:
http://www.ukoln.ac.uk/projects/iemsr/.
[34] IEEE WG12: Learning Object Metadata (LOM) is a
working group focused on the
standardization of learning object metadata:.
http://ltsc.ieee.org/wg12/.
|