(DRAFT COPY: DO NOT QUOTE OR CITE)
Senior Research Scientist
OCLC Online Computer Library Center, Inc.
Office of Research
6565 Frantz Rd.
Dublin, Ohio, 43017, USA
Project Leader, Digital
Library Research Group
Department of Computer Science, Cornell University
Ithaca, NY 14853, USA
The Dublin Core Metadata Workshop Series began in 1995 with an invitational workshop intended to bring together librarians, digital library researchers, content experts, and text-markup experts to promote better description standards for electronic resources. The Dublin Core is a 15-element set of descriptors that has emerged from this effort in interdisciplinary and international consensus building. This paper summarizes the state of the ongoing Dublin Core effort as of January, 1997. It describes the design goals that motivate the Dublin Core, summarizes the results of each of the workshops held thus far, provides brief descriptions of pilot projects that use the Dublin Core, and points to anticipated future developments.
Keywords: Dublin Core, Resource Description, Metadata Warwick Framework
The World Wide Web has revolutionized both the distribution and retrieval of information. For many communities, it has already become the preeminent medium for dissemination of information, and there is no abatement of growth on the horizon.
The rapid expansion in the number of networked resources has not been matched by a commensurate improvement in the technology to aid in the discovery those resources. To this date, virtually all solutions to the resource discovery problem are variations on web content indexing. This model depends on periodic visits of all networked resources and indexing of the full content of these resources. While these web indexes have proven quite useful, the content indexing model is flawed in a number of ways. It is bandwidth-intensive, relies on unrestricted access to resources, and is limited to textual resources.
Rather than viewing web content indexing as the solution to networked resource discovery, we view it as a compliment to indexing based on surrogates of the actual resources; commonly referred to as descriptive or cataloging metadata. The association of standardized descriptive metadata with networked objects has the potential for substantially improving resource discovery capabilities by enabling field-based (e.g., author, title) searches, permitting indexing of non-textual objects, and allowing access to the surrogate content that is distinct from access to the content of the resource itself.
Descriptive cataloging is well-established in traditional libraries, where the MARC (Machine Readable Cataloging) record has been standardized. Maintenance and creation of MARC records requires skilled catalogers and, as a result, the investment per record is high. Such expenditures are inappropriate for many networked resources, which are often more ephemeral than the physical artifacts in libraries and are usually created and maintained by more casual authors or managers.
The Dublin Core Metadata Workshop Series is an ongoing effort to formulate an alternative description standard for networked objects. The goal is to develop a core element set that provides adequate data for Web resource discovery and is simple for authors and content managers to create and maintain. This process has involved a diverse community of librarians, digital library researchers, and text-markup specialists who have convened over four workshops and communicated via an active mailing list. The Dublin Core effort has produced three significant results:
The process of developing consensus among this diverse community is as important a result as the Dublin Core element set itself. This social process is a necessary compliment to the theoretically oriented development of formal ontologies that is currently a prominent line of research in the digital library community. One may think of this workshop series as an attempt to identify an emergent ontology; that is, a consensus among experienced practitioners across many disciplines about the basic elements of resource discovery.
The current state of the Dublin Core - the element definitions, metadata context, and workshop series - is one stage in an ongoing process. As the process continues, we are confident that practical and interoperable tools and technologies for resource discovery will emerge.
The fundamental motivation for the Dublin Core has remained unchanged over the workshop series. That is, the creation of a simple yet effective description element set for a large class of networked resources. Within this rather broad mandate, the focus of the Dublin Core process has evolved as the participants have considered more general metadata issues. The need for a more comprehensive view grew out of a debate with two fundamental dimensions.
The first of these is the scope of the Dublin Core. What are the elements most appropriate for simple resource description? What type of resources is the core meant to describe? How does the Dublin Core relate to other metadata sets? The initial scope of the Dublin Core was intended to be elements to support resource discovery, but it quickly becomes clear that borders between different varieties of metadata (resource discovery and rights management, for example) are less clear than one might hope. Identifying appropriate limits to the scope of the element set has remained a central point of debate throughout the process.
The second dimension of the debate is the degree to which the Dublin Core should be focussed on the Web. While the motivation for the first workshop grew from the rapid proliferation of Web resources, it was quickly acknowledged that these issues will outlive current Web technology and proposed solutions should, as much as possible, accommodate future technical possibilities.
The result of this debate has been the definition of four requirements for a core resource description record:
These requirements work at cross-purposes to some degree, but all are appropriate goals. The ensuing two years of discussion have been, in part, an exercise in minimizing the tensions among them.
It was expected at the outset that authors of resources would construct much of the metadata found on the Web. Thus, elements must be relatively few in number and simple enough to be generated by those unschooled in the cataloging arts. This is perhaps the most tenuous of all the requirements. It remains to be tested whether authors will make the extra effort to attach descriptive metadata to their work, or if that metadata (lacking controlled vocabulary or authority control) will be useful.
Simplicity is a desirable characteristic whether an untrained author or an experienced cataloger creates the metadata. The cost of describing conventional print resources has long been a prominent component of library budgets, and to the extent that simplification leads to lower costs (i.e., more resources described per unit of cost), it is a worthy objective.
It may be argued that simplicity is no bargain if it compromises discovery. Decades of experience with richer description sets should not be sacrificed on the alter of cost containment. Different communities and user groups have different needs, and no single description model is likely to fit all of these needs. One can think of resource description as an ecology of niches, each of which contributes to the overall discovery environment in different, often complementary ways. Thus, full-text Web indexes have the virtue of low per-record expense, which is a result of the automatic nature of their creation. On the other extreme, a MARC record, the lingua franca of resource description in the library world, is likely to be a better discovery surrogate for a resource, but requires cataloging expertise to create and maintain, making it a costlier record to produce and manage. The Dublin Core is intended to fill a niche somewhere between these extremes.
One of the consequences of the networked world is that the artifacts of disparate information domains are pooled in the commons of the Internet. Discovery of information within this commons suffers from the confusion caused by different terminology and description models. Users will benefit from convergence on a domain-independent resource discovery model. This is not to say that there are no inherent differences among disciplines, but rather that where there is a common core of elements, communities should adopt common terminology. The result will be an increase in the visibility and accessibility of their resources to those outside the immediate boundaries of their discipline.
A useful metaphor for this problem arose from the Image Metadata Workshop in this series. What became known as the "digital tourist metaphor" (IMAGE-META-SUMMARY) suggests that the Dublin Core can serve as a kind of constrained discovery vocabulary for finding one's way in the digital world, just as a tourist might use a phrase book or pocket dictionary to accomplish simple communication in an unfamiliar culture.
Full accessibility to the culture and its services requires mastery of the vocabulary of the local environment, whether it be Paris or the discipline of art history. However, a simple resource description set such as the Dublin Core can provide wider accessibility to many resources that might otherwise escape the attention of the non-specialist searcher.
Internationalization is among the principal challenges in achieving the global information infrastructure that the Internet promises. The World Wide Web has lowered barriers to this goal, but while the technology is ubiquitous, the social reality of the Web is still "English Spoken Here".
Resource description suffers from the same linguistic parochialism as the rest of the Web, and the Dublin Core effort up to this time has done little to alleviate this problem. This is not due to a lack of recognition of the problem, but rather to a need to partition objectives into achievable subtasks. As a precursor to undertaking this difficult problem, the organizers have made a concerted effort to involve experts from many countries in the formation and evolution of the emerging consensus.
Participants in the Dublin Core effort have encountered the tension between maintaining flexibility and extensibility on the one hand while containing the element set within a reasonable semantic scope on the other. No discrete set of elements will satisfy all resource discovery needs, and any attempt to create and evolve a monolithic resource description set is unlikely to succeed. The more promising alternative is to promote the evolution of modular resource description packages, administered by separate "communities of expertise", subject to different security paradigms, and expressed via different syntaxes. This is the approach suggested by the Warwick Framework, described in a later section.
Yet, even with the modular approach, we recognize that a mechanism to support extensions to a core package must be defined. This will provide a mechanism for adopting a core set even if the data managers judge that it is incomplete for their local purposes and it also provides a means for experimenting with new elements that might at some point be promoted to core status.
The Workshop Series brings together librarians, digital library researchers, networking professionals, text-markup specialists, and content experts in order to develop an interdisciplinary consensus on the development of common models for network resource description. Significant effort has been invested in ensuring international participation, the result being attendance by conferees from North America, the UK, France, Germany, The Netherlands, Greece, Sweden, Norway, Finland, and Australia. The Fourth Workshop, planned for March of 1997 in Canberra, Australia (DC4-HP), attempts to further broaden participation to Asian countries. Registrants include librarians, researchers, and practitioners from 10 countries in North America, Europe, Asia, and Australia.
The first workshop (DC1-HP) resulted in the original 13-element Dublin Core, a simple element set for discovery of networked resources. The scope of the workshop was limited to the semantics of elements to support discovery of document-like objects (DLOs), putting aside other important varieties of metadata and issues of encoding syntax.
The primary impact of this meeting was to raise the awareness of common interests among the various stakeholders, and to anchor continued development of a common description model for electronic resources. The precise character of the resulting element set was perhaps less important than the consensus achieved among the various communities represented (DC1-REPORT (DC1-DLIB). This consensus has been sustained to a substantial degree in the course of the succeeding workshops, with only modest subsequent embellishments to the element set.
The second workshop (DC2-HP) included consideration of syntactical issues for deployment of the Dublin Core and the development of guidelines for application. The original choice not to address syntax issues in the first workshop was critical to making progress on semantic issues, but left the difficult problem of identifying a practical implementation path, including an encoding and transport syntax that could be deployed in the existing networking infrastructure.
While the Web is not the only environment requiring improved resource discovery facilities, it certainly must be looked upon as the most strategic application. A subcommittee of the workshop adopted the task of defining a syntax for the Dublin Core that is expressible in the formal language of SGML Document Type Definitions, as well as a secondary version that is designed for the constraints of HTML (DC-SYNTAX). The principal challenge of this task is to do so within the capabilities of existing software, i.e., web browsers and HTML parsers.
The effort to develop an HTML encoding for metadata moved forward in a workshop held in May of 1996 by the W3C (The Distributed Indexing and Searching Workshop) (HTML-META). Several projects have taken advantage of this consensus to develop tools that assist in the encoding and embedding of Dublin Core Metadata in HTML (MY-META), (UK-META).
Perhaps the most significant advance of the Warwick Workshop was the discussion of a metadata architecture that placed resource description metadata in the larger context of the many varieties of metadata. This Warwick Framework is more fully described later in this paper.
The final accomplishment of the Warwick meeting was the formation of a group to write user guides for the application of the Dublin Core. It fell to this group to develop recommendations about how to write metadata. Close scrutiny of the elements and their intended meanings unveils details that must be worked out over time and incorporated into recommendations of best practice. The questions raised in this discussion, as well as recommendations from the third workshop, led to modifications of the element set and some of the names of the elements (see table 1). The details of application are far from settled, but the basic 15-element set agreed upon in December of 1996 is stable for the foreseeable future (DC-REF).
The Image Metadata Workshop had the objective of identifying a set of descriptors for images that parallel those of the original Dublin Core as intended for document-like objects. Seventy practitioners in the area of networked image description convened to explore the usefulness of extending the Dublin Core metadata element set to the domain of digital images (DC3-HP).
The workshop focused on description of visual resources such as photographs, slides, and image files. Other formats (motion pictures or video clips, for example) also fall in this category to the extent that they are discrete and bounded. The description of databases or applications having visual outputs of a dynamic character (see, for example, the Visual Human Program at the National Library of Medicine VISIBLE) were judged sufficiently different to require separate consideration.
The workshop consensus supported the notion that the Dublin Core, within the context of the Warwick Framework, affords a foundation for the development of a simple resource description model to support network-based discovery of images (items or collections of items, online or offline). The workshop acknowledged the benefit of a common set of elements supporting discovery of textual and visual resources, and proposed refinement of the Dublin Core element set to better accommodate the description of visual resources (IMAGE-META-SUMMARY), (DC3-DLIB).
The next in the series of workshops will be held at the National Library of Australia in March of 1996 (DC4-HP). Cosponsored by the National Library of Australia,the Distributed Systems Technology Centre (DSTC), and OCLC, the goal of the workshop is to carry forward the common model of networked resource description that has emerged from the first three workshops, and resolve some of the remaining open issues that impede wider deployment.
Many workshop participants have remained active and engaged in discussions concerning the advancement and refinement of this model. Others who have not attended the workshops have also joined this online discussion and have made substantive contributions to progress. These discussions have been an integral part of moving the agenda forward and resolving problems concerning the element set. The META2 mailing list is hosted by Knight and Hamilton at Loughborough University in the UK (META2).
The first workshop resulted in a preliminary core set of elements for description. Thirteen in number, this set represented the consensus of participants on the basic elements for discovery of document-like objects. As a result of subsequent workshops and extensive discussions on the META2 mailing list, the names of some of the elements were changed (to make them somewhat less text-centric and two additional elements were added).
Table 1 describes the elements of the set as currently constituted (DC-REF) for the reference description of the Dublin Core elements). No further changes in the basic set are expected in the near future, although application of some of the elements will be subject to further specification and refinement.
The reference description of the elements resides at http://purl.org/metadata/dublin_core_elements.
The organizers of the 1995 Dublin Metadata Workshop intentionally limited its scope in order to produce concrete results. It became obvious by the second workshop that a broader perspective was necessary in order to formulate answers to three fundamental questions:
A thorough examination of these questions requires a wider perspective that considers the broader characteristics of metadata. Relevant observations include:
Starting from this broader perspective, a number of the participants at the second Workshop formulated and elaborated an architecture, the Warwick Framework (WF), for aggregating multiple sets of metadata.
The Warwick Framework has two fundamental components. A container is the unit for aggregating typed metadata sets, which are known as packages. Independent of the implementation, the only operation defined for a Warwick Framework container is one that returns a sequence of packages in the container. There is no provision in this operation for ordering the members of this sequence and thus no intrinsic way for a client to assume that one package is more significant than another.
At the container level each package is an opaque bit stream. One implication of these properties is that any encoding (transfer syntax) for a container must allow the recipient of the container to skip over unknown packages within the container. This property also permits the contents of individual packages to be encrypted, permitting the transport of metadata across systems that need not have access to specific sets, or that may need to acquire (i.e. purchase) such access.
Each package is a typed object; its type may be determined after access by a client or agent. Packages have three major types, with the possibility that the architecture may be extended through sub-typing.
Figure 1 illustrates a simple example of a Warwick Framework container. The container in this example is a wrapper for three logical packages of metadata. The first two, a Dublin Core record and a MARC record, are internal to the container. The third metadata set, which defines the terms and conditions for access to the content object, is referenced indirectly via a URI in the container.
The Warwick Framework may be implemented, with varying degrees of power, using a number of techniques. Some that have been proposed or planned are HTML, MIME, SGML, and using a distributed object architecture such as CORBA.
The mechanisms for associating a Warwick Framework container with a content object depend on the implementation of the Framework. For example, with no extension to the current HTML syntax, a Warwick Framework container can be embedded in an HTML document using the META tag (HTML-META). Simple indirect linkage to a component package can be done using the LINK tag. Other proposed implementations, which are not constrained by the current Web architecture, have more flexible and powerful metadata association mechanisms.
The reverse linkage, that which ties a container to a piece of intellectual content, is also relevant. Anyone can, in fact, create descriptive data for a networked resource, without permission or knowledge of the owner or manager of that resource. This metadata is fundamentally different from that metadata that the owner of a resource chooses to link or embed with the resource.
We therefore informally distinguish between two categories of metadata containers:
Figure 2 illustrates the relationship between two externally-referenced metadata containers and a content object that includes an internally-referenced metadata container. Note that all three objects have uniform resource identifiers (URI's).
There are now major projects in the US, Europe, and Australia that make use of the Dublin Core element set. Short summaries of some of these projects are included below.
The Solinet Monticello Electronic Library Project (MONTICELLO) is designed to link distributed regional resources regardless of source or type of information. This project focuses on two approaches to accomplish this goal. The first is a high-level semantic description based on the Dublin Core Element Set to provide predictable access points into differing collections of information; the second is the application of HTTP-Z39.50 servers to allow for the integration of distributed collections of information for effective retrieval. These distributed collections include Government Information Locator Service (GILS), the Encoded Archival Description (EAD) for special collection finding aids and Machine Readable Cataloging (MARC) records.
The Distributed Systems Technology Centre (DSTC) in Australia is working with the Dublin Core model on a number of projects (DSTC):
The Nordic Metadata Project (NORDIC) is driven by the high demand for regional interlibrary loan and resource sharing. This demand has resulted in the creation of a Nordic metadata production, indexing and retrieval environment. The system serves the dual purposes of experimentation and production-quality discovery and retrieval for end-users.
Project team members have been active in the development of the Dublin Core specification and are working on the important issue of cross-walks between Dublin Core and Nordic MARC formats. The active participation of this group in the development of indexing robots will contribute to integrating the Dublin Core into existing infrastructure.
The German Mathematical Society provides WWW access to its preprint collection using Harvest (MATH-PREPRINT) (HARVEST). Most of these preprints are stored in Postscript format, but each has an HTML "shadow file"(essentially a title page with abstract and metadata). They have adopted the convention for embedding metadata elaborated at the W3C workshop on distributed indexing and searching (HTML-META) and are in the process of making their metatags compliant with the recently stabilized 15-element Dublin Core.
Part of this effort includes a form-based page (My-Meta-Maker 1.2) to allow authors to fill in element attributes that are subsequently formatted appropriately for inclusion in an HTML document.
Automated or semi-automated tools for capturing and structuring metadata are on the minds of others as well. Paul Miller and Tony McDonald have developed a simple form for embedding basic Dublin Core metadata in Web pages (UK-META). Initially prototyped for the original Dublin Core element set, it has been modified to support the 15-element set. It is expected to be adopted by the Nordic Metadata Project team for use by the Nordic project sites, in addition to the Arts and Humanities Data Service for which it was developed.
Ariadne (ARIADNE) is the online digital library journal of the United Kingdom Office for Library Networking (UKOLN) Beginning in January of 1997, all the articles in each issue of Ariadne will contain Dublin Core metadata. Back issues (6 back to 1) will be retroconverted as well, making Ariadne the first Dublin Core compliant electronic journal.
NewsAgent for Libraries (NEWS) is a two year project in the UK Electronic Libraries Programme (eLib). The aim is to create an electronic news and current awareness service with a mixture of content, including descriptions of documents. The intended users are library and information staff. Content will include refereed papers as well as reviews and editorial matter from the UK journals in the field. News and briefing materials will also be provided by organizations in the field such as The British Library, LITC and the Library Association.
UKOLN's involvement is in writing software to enable the NewsAgent server to harvest metadata (in the form of embedded Dublin Core) from various information sources. Content providers will be requested to embed Dublin Core descriptors in their Web pages, and this metadata will be harvested to update a central database. Additional attributes will be used internally, e.g. to hold copyright information.
ROADS(ROADS) is a UK eLib-funded project whose aim is to develop discovery software for internet resources. ROADS will be suitable for a number of directory applications (research registers, project databases, and similar applications), but primarily it is being developed to fulfill the needs of the eLib subject-based services, which feed their requirements into the project.
One of the aims of the EC DESIRE project is to develop subject-based internet resource discovery services in a European context. As part of this effort, the ROADS software will be extended to enable harvesting of embedded metadata. Plans include using harvesting techniques to enable details of individual resources to be preloaded during manual record creation. Also, bulk creation of records will be accomplished by harvesting from a list of pre-defined high-quality `root' URLs. It is likely that several versions of the harvesting software will be supplied, based on different third-party software packages such as the Harvest suite.
The intention is to harvest embedded Dublin Core metadata where possible. In the particular case of bulk-harvesting document-level resources from lists of service-level resources it is hoped to tie the resources together by using the Dublin Core RELATION attribute.
ADAM is an information gateway to quality-assured resources on the Internet in art, design, architecture and media. The Dublin Core is used in this project as the content description standard.
The Visual Arts Data Service (VADS), a new agency of the Arts & Humanities Data Service, will provide the UK Higher Education community with Dublin Core-mediated access to networked electronic research data in the visual arts, following consultation with relevant constituencies regarding the development, promotion, and implementation of standards for best practice in data creation, collection, description, delivery and preservation.
The Z39.50 Profile for Simple Distributed Search and Ranked Retrieval, ZDSR, will employ the Dublin Core elements. ZDSR originally was based on the Stanford Protocol for Internet Search and Retrieval (STARTS), an initiative of the Stanford Digital Library Project, which developed requirements for distributed searching and ranked retrieval during the Spring and Summer of 1996.
In the summer of 1996, Z39.50 implementers began collaboration with participants in the STARTS project to develop ZDSR, originally called the ZSTARTS profile (Z39.50 Profile for STARTS). The profile assumes that queries pertain to text documents and that retrieval records consist of document metadata. Thus, a client searches for documents, and retrieves document descriptors.
The profile supports searching by various fields as well as by full-text keywords. It also supports relevance feedback searches, and stem and phonetic searching. Search results may be restricted by threshold score, or by maximum number of documents. The query includes both a restriction component and a ranking component. The restriction component is the normal Z39.50 Boolean query, used to specify the documents that qualify for subsequent ranking. The ranking component is a list of terms, each assigned a relative weight by the client.
We are encouraged that there has been sufficient progress on the Dublin Core to warrant the deployment efforts described above. There is no doubt, however, that there is ample opportunity for refinement of the specification of the Dublin Core and Warwick Framework. In this section, we describe some of the open issues and a number of ongoing development efforts.
The 15 element set is now established, but specifics of how these elements should be applied is less clear, requiring further discussion informed by experience and early prototyping efforts. The following issues are prominent on the current agenda of Dublin Core issues:
The embedding of metadata in HTML described, in a previous section, provides a way to deploy metadata in current documents without compromising existing markup standards or breaking servers and clients. Wide deployment of metadata will require more sophisticated means of encoding and transferring such data.
Rather than develop a new encoding scheme from scratch, it makes sense to leverage related work. This opportunity exists with PICS (Platform for Internet Content Selection) standard, which has been developed under the auspices of the W3C in response to the need to provide third-party content rating schemes for Internet resources (PICS-DESC).
A meeting of the PICS working group in London in January of 1997 produced proposals to modify the standard to make it sufficiently flexible to accommodate general resource description. If these modifications develop as now expected, the PICS scheme may provide an important link in the deployment of better resource description systems, including widespread support for metadata in browsers, servers, and authoring systems (PICS-NOTES).
In a world that embraces many data content standards, it will often be necessary to map the relationships and equivalencies among these systems. Efforts are under way to do this for the Dublin Core and related content standards. The Library of Congress has recently made available Discussion Paper 99(LC-99) and a paper on a Dublin Core/MARC crosswalk which address cross-mapping issues for the Dublin Core and MARC records.
UKOLN supports a Web site that provides interoperability information for various content standards(UK-MAPPING). Further information about related resource description can also be found in the survey of formats by Dempsey and Heery (DESIRE-META)
The Warwick Framework provides the conceptual structure for aggregating physically distinct metadata sets. There are a number of topics that call for more detailed and extended examination prior to full implementation of the framework.
The most fundamental question about the Warwick Framework is the semantic interaction among the multiple packages in a container and between those packages and content objects. Abstractly, we can think of these semantic interactions as a set of relationships, with metadata sets and content objects playing roles in these relationships(WF-DANIEL). Two established technologies provide inspiration for this conceptualization of the problem:
Using the concept of relationships, one can imagine a separate package or a header record in a Warwick Framework container that acts as a catalog of the relationships that express the semantics of the container. When a client or agent first opens a Warwick Framework container, it might read the catalog. This would then establish a schema with which it could then interpret the remainder of the catalog. Defining the structure of this catalog and the use of it by clients and agents is currently being investigated.
The Dublin Core Metadata Workshop Series is nearing the end of two years of effort. On several levels it has been a great success. It has provided the context for several communities to work together and share expertise on the problem of improving resource discovery in a networked environment. There is now agreement on the composition and semantics of a core descriptive metadata set. There is an outline of a framework that provides a broader metadata context for this core descriptive metadata set.
The fact remains that there is still very little metadata out there on the Internet. Resource discovery remains one of the fundamental problems of the information infrastructure. The quantity of networked information is growing exponentially, while the tools for locating and accessing that information lag behind.
The challenge for the next stage of the Dublin Core effort is to move beyond ideological consensus and cope with some of the implementation issues in deploying networked descriptive metadata. This must be done in a manner that integrates technical needs on the one hand with the social behaviors of both information managers and end-users on the other hand. This process has already begun and will continue with some of the implementation efforts that have been described in this paper. In conjunction with related efforts to further define the engineering issues associated with the Dublin Core, we are optimistic that network resource discovery will progress beyond its current immature phase.
Notes on the January, 1997 W3C PICS Meeting, Paul Resnick, January, 1997, http://www.research.att.com/~presnick/pics/0196wg/Overview.htm
PICS Label Distribution Label Syntax: Version 1.1, Tim Krauskopf and Jim Miller and Paul Resnick and Win Treese, W3C, http://www.w3.org/pub/WWW/TR/PR-PICS-labels-960808.html
Metadata, Dublin Core and USMARC: a Review of Current Efforts, Library of Congress , January ,1997, MARBI Discussion Paper 99, gopher://marvel.loc.gov/00/.listarch/usmarc/dp99.doc
Metadata: Mapping between Metadata Formats, http://www.ukoln.ac.uk/metadata/interoperability/
Work on relationships and Warwick Framework catalogs is joint with Ron Daniel Jr., Advanced Computer Lab, Los Alamos National Laboratory.
The ROADS Project: Resource Organisation and Discovery in Subject-based Services, December, 1996, http://ukoln.bath.ac.uk/roads/
The Warwick Metadata Workshop: A Framework for the Deployment of Resource Description, Lorcan Dempsey and Stuart L. Weibel, DLib Magazine, July, 1996, http://www.dlib.org/dlib/july96/07weibel.html
OCLC/NCSA Metadata Workshop, March, 1995, http://purl.org/oclc/rsch/metadataI
UKOLN/OCLC Metadata Workshop, 1996, http://purl.org/oclc/rsch/metadataII
CNI/OCLC Image Metadata Workshop, September, 1996, http://purl.org/metadata/image
The Fourth Dublin Core Metadata Workshop, January, http://www.dstc.edu.au/DC4
CNI/OCLC Workshop on Metadata for Networked Images: Executive Summary, September, 1996, http://www.oclc.org:5046/research/dublin_core/summary.html
OCLC/NCSA Metadata Workshop Report, Stuart Weibel and Jean Godby and Eric Miller and Ron Daniel Jr., OCLC Online Computer Library, Inc., June, 1995, http://purl.org/metadata/dublin_core_report
Dublin Core Qualifiers, Jon Knight and Martin Hamilton, Department of Computer Science, Loughborough University, January, 1997, http://www.roads.lut.ac.uk/Metadata/DC-Qualifiers.html
CORBAservices: Common Object Services Specification, Object Management Group, Chapter 9, March, 1995.
The Knowledge Interchange Format, Michael R. Genesereth and Richard E. Fikes, Computer Science Department, Stanford University, June, 1992, Logic Group Report Logic-92-1
Metadata: a survey of current resource description formats, Lorcan Dempsey and Rachel Heery, UKOLN, October, 1996, Draft Version of Work Package 3 of Telematics for Research project DESIRE (no. 1004)
Metadata: The Foundations of Resource Description, Stuart L. Weibel, DLib Magazine, July, 1995, http://www.cnri.reston.va.us/home/dlib/July95/07weibel.html
Image Description on the Internet: A Summary of the CNI/OCLC Image Metadata Workshop, Stuart L. Weibel and Eric Miller, DLib Magazine, January, 1997, http://www.dlib.org/dlib/January97/oclc/01weibel.html
A Syntax for Dublin Core Metadata: Recommendations from the Second Metadata Worshop, Lou Burnard and Eric Miller and Liam Quin and C.M. Sperberg-McQueen, July, 1996, http://users.ox.ac.uk/~lou/wip/metadata.syntax.html
Dublin Core Reference Description, January, 1997, http://purl.org/metadata/dublin_core_elements
A Proposed Convention for Embedding Metadata in HTML, June, 1996, http://www.oclc.org:5046/~weibel/html-meta.html
My-Meta-Maker 1.2, http://www.mathematik.uni-osnabrueck.de/projects/META
The Dublin Core Metadata Creator, http://www.ncl.ac.uk/~napm1/dublin_core/
The Visible Human Project, http://www.nlm.nih.gov/publications/factsheets/visible_human.html
Monticello Project, http://www.solinet.net/monticello/monticel.htm
The Distributed Systems Technology Centre (DSTC) Metadata Activities, http://www.dstc.edu.au/RDU/
Nordic Metadata Project, http://linnea.helsinki.fi/meta/index.html
German Mathematical Society Preprint Project, http://www.mathematik.uni-osnabrueck.de/ak-technik/anlagen/goevt.html
The Harvest Information Discovery and Access System, http://harvest.transarc.com
Ariadne, An Electronic Journal published by the United Kingdom Office for Library Information and Networking (UKOLN), http://www.ukoln.ac.uk/ariadne/
DESIRE: Development of a European Service for Information on Research and Education, http://www.nic.surfnet.nl/surfnet/projects/desire
NewsAgent, http://www.ukoln.ac.uk/metadata/NewsAgent/dcusage
Art, Design, Architecture & Media Information Gateway, http://adam.ac.uk/
The Visual Arts Data Service, http://vads.ahds.ac.uk/
META2 mailing list, managed by Jon Knight and Martin Hamilton at Loughborough University in the UK
The Warwick Framework: A Container Architecture for Aggregating Set of Metadata, Carl Lagoze, Clifford A. Lynch, and Ron Daniel Jr., Cornell Computer Science Technical Report TR96-1593, July, 1996, http://cs-tr.cs.cornell.edu:80/Dienst/UI/2.0/Describe/ncstrl.cornell/TR96-1593
The authors extend grateful appreciation to Terry Allen of Fujitsu, Mark Bendig of the OCLC Office of Research, Ron Daniel of Los Alamos National Laboratory, and John Kunze of the University of California at San Francisco for substantive editorial comments that improved this manuscript.
The authors are also indebted to the following for information on projects with which they are involved: Ray Denenberg of the Library of Congress, Tony Gill of the Surrey Institute of Art and Design, Rebecca Guenther of the Library of Congress, Rachel Heery of the United Kingdom Office for Library and Information Networking, Renato Ianella of the Distributed Systems Technology Centre, Traugott Koch of Lund University, Eric Miller of the OCLC Office of Research, and Paul Miller of the University of Newcastle.