Metadata Summit Summary

This meeting on metadata was originally proposed by the Task Force on Meta Access, sponsored by ALCTS (Association of Library Collections & Technical Services, a division of the American Library Association). RLG, through a group of seven meeting planners (see the summit participants list), brought together knowledgeable representatives from a variety of research information projects, standards groups, and related communities.

These participants represented information providers and those responsible for making the provided information intelligible to end users. The summit's intent was to focus on using metadata to improve discovery of what are now referred to as “deep Web” resources. It was held on July 1, 1997, at RLG headquarters in Mountain View, California.

Presentations

Two brief introductory presentations emphasizing the Dublin Core metadata elements and their application helped refine the focus of the meeting:

1. Stuart Weibel, OCLC, summarized the evolution of the Dublin Core. He reviewed current developments that allow Dublin Core implementation in the Web environment. His presentation underlined the continuing impetus fueling development of the Dublin Core, its growing international acceptance, and the fact that its development has now reached a stage that permits implementation in Web applications.

2. Ricky Erway, RLG, extended the discussion of metadata to the issues involved in providing access to deep Web information. She discussed the confusion inherent in the mixed focus of metadata on the "Webness" of an entity and on the entity itself. She pointed out that this issue is what makes use of the Dublin Core especially problematic for the research community; it needs to be sorted out if use of the Dublin Core is to be extended beyond application to Web pages.

Discussion

Underlying all aspects of the discussion was the need to be user-centric. Recognizing that users differ in their needs, information-seeking goals, and searching sophistication, it was noted that when domain-specific resources are accessed by users other than those for whom they were originally designed, the differences are multiplied.

While the group endorsed the strategic application of the Dublin Core elements for annotating static HTML Web pages to make indexing and searching of Web pages more effective, they emphasized extending the use of the Dublin Core and finding additional ways to use it to help researchers discover a wider range of non-HTML based information. Examples included electronic text centers, library catalogs, and "real-world" collections that may become known to users via Internet-accessible information.

The participants acknowledged that only in some cases would people use the Dublin Core elements for cataloging. Rather, the elements might function as user access points to search metadata, or as common denominators or indexes among existing, diverse metadata schemes. Alternatively, one might use the Dublin Core elements to identify resources that could profitably be further investigated in their native form.

Issues of granularity and of descriptive perspective came into play. How, for instance, could users "unfold" various levels of representation (collection, series, or item, for example) in search results if they were combined indiscriminately? And how could metadata be made to represent the difference in the relationship between a book, an electronic copy of the book, and images of the pages of the book?

There was agreement that an important function of the Dublin Core might be to provide a locus for the mapping of data elements from heterogeneous metadata systems, to facilitate interoperability.

Additionally, the Dublin Core could provide a "window" to specific data resources that use other descriptive metadata models. Once found, the user would then query that database in its native form.

This function of the Dublin Core would approximate that of a directory of directories serving to get one to a specific location within which one might search for objects of interest.

Much of the discussion centered on making the Dublin Core elements fit "our" stuff—that is, primary or secondary research materials that may also be represented by digital reproductions, or may be represented by Internet-accessible citations or bibliographic records. The Dublin Core elements were designed to describe Web documents. Consequently, they do not distinguish a relationship between object and source since they pertain to the "item" only as it is represented in a Web page. The frequently suggested solution of using the SOURCE element for all information about the original item was insufficient to resolve the confusion effectively. Lumping all information about the original into SOURCE would not be helpful to researchers when they are concerned about the "thing" rather than with the Web page representing it.

Internet access to a range of non-HTML based information resources necessitates navigation of a heterogeneous, distributed environment and implies interoperability between metadata schemes. Consequently, much of the discussion centered on mappings, or crosswalks, between domains.

Participants agreed that it is important that descriptions of metadata systems be available and maintained in some central, authoritative location. Consensually developed metadata systems eventually become de facto standards, particularly when they are reliably maintained and can "piggy-back" on other standards.

Conclusions

Participants agreed that the communities represented at the summit should:

• Support the continuing development of the Dublin Core strategic Web application.
• Encourage widespread experimentation with implementations using the Dublin Core elements.
• Encourage the production of user guidelines specific to particular domains that will be publicized and shared.
• Defer public reporting of non-HTML based uses of the Dublin Core elements, since to advocate other uses prematurely risks creating confusion.

They affirmed that:

• It is too early to attempt to institute formal standards processes.
• Each community can provide an appropriate venue for informally vetting the efficacy of particular metadata implementations, mappings, etc.
• The Dublin Core, because of the increasing buy-in of many diverse communities, offers a convenient organizing point for "translating" between metadata systems.
• The Dublin Core is still in an early stage in its development and it is important to participate in its continuing articulation if it is effectively to meet the needs of the research community, particularly in regard to problematic definitions of certain elements.
• Even though formal standards processes are not necessary, it is important that de facto standards, such as the Dublin Core, be made available and maintained in a stable location.

Subsequent working group activity

Summit participants formed a working group to review the Dublin Core in terms of the research community's needs, provide a non-HTML perspective to implementation, and investigate problems intrinsic to specific elements. This group sent its proposed guidelines to the DC5 meeting in Helsinki in October 1997.