There is now ongoing work on producing a Z39.50 profile for Dublin Core.
The first step in that work is the consolidation of the extended Tagset-G
and Tagset-M. For those who are not into the Z39.50 jargon (I'm almost a
newbie myself) I would describe the tagsets as the basic "parts of speech"
in the formulation of a data model for use in Z39.50. In addition to
these tagsets there are other that are specific to given profiles and
there are even tagsets specific to given applications.
In principle, the tags are used to define a hierarchical field
structure which fits very well into the dot model for semantic
refinement discussed here within Dublin Core initiative, as for
example (taken from Rebecca's subelements page
http://www.loc.gov/marc/dcqualif.html)
DC.Publisher
DC.Publisher.PersonalName
DC.Publisher.CorporateName
DC.Publisher.PersonalName.Address
DC.Publisher.CorporateName.Address
The issue I want to raise is: How should we understand the subelements
when we are searching? Consider the following two pseudo-SGML
examples.
<publisher>
<CorporateName>The globe theater</CorporateName>
<CorporateName.Adress> Somewhere in
London</CorporateName.Adress>
</publisher>
and
<publisher>
<type> CorporateName
<value> The globe theater
</publisher>
<publisher>
<type> CorporateName.Adress
<value> Somewhere in London
<scheme> Postal
</publisher>
The first example implies that we may want to search for 'London' in
the the database's publisher.CorporateName.Adress field (subelements
are thus to be regarded as refined _fields_). The latter case means
that we want to search for 'London' in the publisher.value field *AND*
for 'CorporateName.Adress' in the publisher.type field (where in
effect subelements are taken from a _vocabulary_ of refinements)
Assuming that Alta Vista would support dublin core, one would enter
either of the following in the advanced form
publisher.CorporateName.Adress:London
or
publisher.value:London AND publisher.type:CorporateName.Adress
In fact Z39.50 may allow us to do something better than a logical AND
in the last example (a proximity where we are in effect looking at who
is sitting on the twig next to us in the GRS tree)..
It seems to me that type and scheme qualifiers are refining the
semantics of an attribute and clarifying its a value respectively.
Somewhat like a physical quantity (speed, say) and its dimension
(meters per second).
I think we should give input to the ZIG as to how we want the 15
elements to be handled. In the current tagset proposal each of them
may be regarded as being of a structured datatype, having sub-elements
type and scheme. We would presumably want to add lang.
However, this data model does infact imply that we are thinking of the
types as being taken from a VOCABULARY OF REFINEMENTS. Is that what we
want? Is it this way people are thinking upon subelements in general?
Cheers
Sigfrid
________________
Sigfrid Lundberg, Ph.D., . [log in to unmask]
Lund University Library, http://www.ub.lu.se/~siglun/
Netlab, PO Box 3, S-221 00 Lund phone +46 (0)46 222 36 83
|