The NKOS Special Session at DC2005 was a stimulating session, filled with people who’d thought long and hard about the issues involved in knowledge management. The session, chaired by Traugott Koch (now of UKOLN), brought together enough people and ideas to make a whole afternoon go by very quickly.

The presentations started with Doug Tudhope, of the Hypermedia Research Unit, University of Glamorgan, speaking about “Applications of NKOS: some examples and questions.” In the course of his talk, he made a point I managed to write down: “The cost/benefit issues for KOS applications lie in the granularity of relationships and degree of formalism.” This should probably end up on a large banner of some kind. Rachel Heery of UKOLN spoke on “(Metadata and) Vocabulary Registries.” Her presentation was an excellent summary of the issues involved in vocabulary registries. She spoke of the functions and goals of this type of registry, in particular the opportunities for re-use, intelligent data exchange, and innovative interfacing with applications from other domains. One of the crucial areas now still fuzzy is that around policies: in particular status, rights, persistence, identification, quality, and the all-important standards compliance. Some of the open questions she identified:

• Who owns content? Can it be withdrawn?
• Who is responsible for transforming the content to a machine-readable structure? Who’s responsible for the maintenance of the content in that form?
• Is there commercial motivation for KOS owners to work together?
• Can a single registry deal with more than one underlying model?

She pointed out again what we already know: large vocabularies have a commercial business model, and have issues around IPR and copyright that smaller ones do not, and distributed registry system will have to deal with all of this. She also emphasized that none of the registries built so far have managed to find appropriate incentives for vocabulary owners to maintain their data in registries. [Diane: And the incentives may be different depending on the size and nature of the vocabulary.] Andy Powell made a short but pungent presentation on the burning issue of “(Persistent) Identifiers for Concepts/ terms/relationships.” He identified the functional requirements for persistent identification to be:

• declare and use
• identify uniquely
• attach definitions and relationships
• allow access
• all persistently

Andy’s generic advice about identifiers could perhaps be summed up as “pay attention to standards, but go with the crowd.” He added some specific stuff about identifiers for metadata terms, particularly URI scheme registration:

• registration is important, and helps to ensure uniqueness
• without registration there are no guarantees of uniqueness OR persistence
• identify all terms using URIs from registered schemes
• URIs need to be resolvable, so that “reasoning” can be performed
• Web Architecture talks not about resolution but dereferencing, more than one representation may be available
• only http URIs offer simple, widely deployed dereferencing mechanism

His suggestion is that there needs to be http content negotiation to select between a human and machine-readable representation. To support that, he suggests the use of project and/or service URIs, or xmlns.com (set up for them) or PURLs. These should then be dereferenced via an http 303 redirect to both human-readable and machine-readable information. Eric Childress gave the final presentation before the break, discussing “OCLC and Vocabulary Identifiers.” He described OCLC’s Terminology Services, which include:

• converting, normalizing and adding value to vocabularies
• releasing vocabularies in a web services environment
• making experimental use of INFO:URI

Gail Hodge presented some “Case studies from the NBII and others” after the break, and lead some discussion, using panelists who’d agreed to be available: Dan Brickley, Joseph Busch, Eric Childress, Ron Daniel, Jane Greenberg, Rachel Heery, Diane Hillmann, Gail Hodge, Johannes Keizer, Traugott Koch, Alistair Miles, Andy Powell, Jian Qin, Stuart Sutton, Joseph Tennis and Doug Tudhope. All presentations will eventually be linked from the NKOS website (http://www.ukoln.ac.uk/terminology/events/NKOSatDCMI05.html), not the conference website.