Unrealistic approaches to metadata quality are legion. We have all seen mark-up-everything-in-sight approaches, sat through discussions of element sets that turn into advanced seminars in 19th century German philosophy, ducked as quality-control issues became ammunition in jurisdictional wars, and watched projects sink without trace into a quagmire of notional perfection.

Suppose we wanted to develop a realistic and organized approach to quality? How would we do that? This no-holds-barred, deeply personal, intimate account reveals shocking truths about metadata quality that were 'too much' for a general audience--until now.

Tom Bruce, Director of the Legal Information Institute (LII) at Cornell's Law School, promises an interesting presentation and discussion of metadata quality-control issues. Prior to co-founding the LII (with Peter Martin) in 1992, Mr. Bruce served for several years as Director of Educational Technologies at the Cornell Law School. He is the author of Cello, the first Web browser for Microsoft Windows, and of a variety of other software tools used by the LII and others. As part of his LII activities, Mr. Bruce has consulted on Internet matters for Lexis-Nexis, West Group, IBM, Folio Corporation, and others.

This Talk is based on a chapter, "The Continuum of Quality: Defining, Expressing, Exploiting", authored by Tom and Diane Hillmann in the forthcoming book from ALA Press, Metadata in Practice, edited by Diane I. Hillmann and Elaine l. Westbrooks.

Tom began by stating that the uncontrollable web puts a burden on information professional in specialized communities of interest because:

• Many information resources are not good at helping users retrieve information efficiently;
• People like librarians fail to use information from others; It can be good enough
• Librarians suffer from a negligible altruism problem that tends to over look the fact that quality is expensive; and a "holier than thou" attitude that insists on standards that can not be followed.

Tom iterated that there is a conversation that librarians should NOT be having:

• Build the information system well and no one will question it
• Build gold standards that are too expensive to anyone to use
• Attempt to achieve realistically; everything can not be perfect
• Others (especially creators) can contribute; there just aren't enough librarians
• Examine the defects of metadata when discussing quality

On the other hand, the conversation that librarians SHOULD be having is:

• There should be an inherent quality that is domain and discipline independent
• the creation methodology should be independent;
• Aim higher than information discovery
• final decisions are conditions by economic and managerial realities

Aspects of Quality

The Magnificent 7 were listed and explained in detail with examples:

1. Completeness
2. Accuracy
3. Provenance
4. Conformance to expectations
5. Logical Consistency and coherence
6. Timeliness
7. Accessibility

There are two approaches to creating and assessing metadata: The techies tend to adopt an iterative approach which assumes that data extraction technology will get better and that we will eventually get it right. On the other hand, most librarians are inclined to state that we should try to get it right at the beginning.

Bruce Summarized the key points:

• Quality is best enforced by community;
• Reinforce user confidence in the data (spelling errors are corrosive)
• Figure out which approach is better for you; Do it once and forget about it OR Handle it iteratively

There are 3 levels or tiers of Quality; ranging from the inexpensive and automated to the expensive and expert-opinion-based; (1 is the worst and 3 is the best)

Level 1

• Schema validation - (Defined schema; automatic validation)
• Appropriate namespace declarations
• Each element defined within a namespace; not necessarily machine-readable
• Administrative wrapper present

Level 2

• use Controlled vocabularies that are defined and documented by the community
• Use the full complement of access points
• More Provenance Information

Level 3 (best)

• The intentions of the metadata are clear; Mostly everyone understands why a particular element exists; and application profile exists
• Full Provenance Information; the source of the data is well known and so is its history

Towards the end of Tom's presentation he discussed a variety of scenarios that guide information professional to ask certain types of questions and to find out where to look for that information if the answer is not available. For example, one may ask, "Does the element set completely describe the object?" If the answer is no, then one should look at the application profile and the documentation

Finally Tom stated that cultural change and metadata augmentation will be the only two things that will improve metadata quality in the end.