Greetings,

In conjunction with one of our research projects here at the University of
Illinois we have about 60,000 full-text journal articles in XML format as
provided to us by various sci-tech publishers. We maintain rich and
extensive metadata for each of these articles. We have begun the process of
expressing that metadata using Qualified Dublin Core semantics and RDF, as
described in the latest draft of "Expressing Qualified Dublin Core in RDF"
(http://www.mathematik.uni-osnabrueck.de/projects/dcqual/qual21.3.1/). An
up to date sample of one of our metadata files is available at
http://dli.grainger.uiuc.edu/uiLIb/05_lecuyer_full.met

We would be interested in comments or constructive criticism from members of
this list regarding the way in which we've marked up our metadata in this
sample file. We found it necessary to use additional namespaces and extend
Qualified Dublin Core:

1. We made use of MathML namespace in order to preserve mathematical
information embedded in titles, abstracts, figure captions, etc.

2. In order to preserve richness of metadata we found it necessary to
qualify DC Creator and Contributor. To do this we appropriated semantics
from the now obsolete December 1999 "DC Agent Qualifiers" working draft
(http://www.mailbase.ac.uk/lists/dc-agents/files/wd-agent-qual.html). As a
working hypothesis we assumed that DC Creator and Contributor could contain
nodes of appropriate DC Agent Type (DCAT1), and that these nodes could have
properties of Agent Type, Agent Affiliation, Agent Role, and Agent
Identifier (all optional, all repeatable). We will update this hypothesis
when the next draft of DC Agent Qualifiers is available.

3. We also found it necessary to add several refinements and encodings to
those enumerated in the current Qualified Dublin Core documentation. We
created an RDF Schema for these local extensions (available at
http://dli.grainger.uiuc.edu/uiLib/uiLib.rdfs), trying to stay consistent
with the spirit of the RDF Schema for DCQ included as an appendix in
http://www.mathematik.uni-osnabrueck.de/projects/dcqual/qual21.3.1

4. We also found it necessary to augment the elements included in the DCMI
Type Vocabulary
(http://www.dublincore.org/documents/2000/07/11/dcmi-type-vocabulary/). The
added types are listed in the RDF Schema for our local namespace as
mentioned above. We realize that the DCMI Type working group is working to
define additional DC Types (and/or subtypes) and still debating how
(whether) local communities might extend approved list of DC Types. Again,
we'll come back to this issue when that work is further along.

5. Finally, we use here namespace URIs for DC and DCQ consistent with those
used in the examples in
http://www.mathematik.uni-osnabrueck.de/projects/dcqual/qual21.3.1. We will
update when consensus about namespace URI conventions is reached.

If you see obvious places where we've gone astray please let us know. We
have so far transformed metadata for about 300 articles provided by ACM and
will be moving on next to samples provided by AIP and APS. It has so far
been an interesting and enlightening experience, albeit somewhat painful and
time consuming as well. The most difficulty we have had to date has been
with trying to understand how to implement local extensions to the base set
of DC Qualifiers. In that respect, it might be desirable to consider
supplementing the current draft of Expressing Qualified Dublin Core in RDF
with another document advising local communities as to how they might go
about creating and implementing their own local extensions where necessary
(and how they might propose additional qualifiers of widespread interest).

Thanks,

Tim Cole
Tom Habing
University Library
University of Illinois at Urbana-Champaign