Metadata Workshop, Luxembourg - 1-2 December 1997

Workshop Report - 8. Breakout session: Retrieval

This was a diverse group, coming from different sectors (libraries, museums, technologists) and having different interests in the types of data and resource to be retrieved (images, cultural resources, bibliographic services and data).

The topics that had been identified in advance and presented to the group as a baseline for discussion were:

• Precision vs. recall (full-text search versus metadata searching, use of sub-elements, use of schemes)
• Cross-domain issues (searching different sets of metadata, relation to Z39.50)
• Multilingual searching (user interface issue or harvesting issue?)
• Searching for collections of documents (levels of metadata)
• User guides/explain information (instructions for use, explanation of results)

In the event, most of the discussion revolved around various aspects of cross-domain searching, including 'levels' of metadata, and on user and usability issues. This did not mean that the other issues were seen as unimportant: multilingualism, in particular, was recognised by all as a critical topic for the future.

Early discussion focused on the need for and relevance of metadata, particularly as defined in the Dublin Core. A number of points emerged:

• concern for the potential for conflicts between existing schemes and solutions for interoperability in retrieval and the DC solution;
• recognition of the complexity of the domain;
• the present experience and application of DC are heavily rooted in the HTML environment (where a syntax for it exists) and this may be a limitation.

Overall, the group agreed that a common approach to metadata was necessary, especially to support cross-domain searching, across different types of objects and descriptive practices. However, in that context, it may be that even a simpler or 'shallower' set of descriptive elements is needed to answer simply the questions who, what, where and how. The model emerging in the discussions was one of essential metadata at a high level of extrapolation with drill-down features to richer and more specific schemes. On the specific issues of Z39.50, the existing Bib-1 attribute set and the relationships to DC, opinions were divided about whether DC should be accommodated within existing Bib-1 mapping or whether a separate DC attribute set should be established.

The second main focus to emerge in the discussion was usability, with the main points being:

• lack of information and knowledge of user behaviour in searching;
• a corresponding lack of understanding and knowledge of user expectations.

As a result, there is an identified need for systems to explain the scope of searches, the context for search results, and the nature of the objects and resources retrieved. More information and understanding is needed of the different layers in the information model and on their impact and use in the retrieval process.

A number of other topics were also identified as important for effective retrieval and delivery systems and which need to be addressed through the provision of metadata, though they fell outside the direct remit of this breakout session. They included the need for stable, authoritative and unique identifiers for objects; and the problems and information associated with rights handling and conditions of access.

[Table of Contents]