Indeed, the experience of trying to build services gives us valuable
information about _what_ is "right" as far as the underlying data goes.
Not that we should design our data for any _particular_ application, but
only by trying lots of applications can we abstract out the general
rules and principles and features of 'good data', informed by actual
practice.
meta-data and the application that uses it; theory and practice. Both of
these pairs must be mutually informed.
Jonathan
Art Rhyno wrote:
>> I think there is a lot of growth opportunity, but libraries have to get
>> the underlying data structures and standards right and then make them
>> available not only for their own internal purposes but for the larger
>> public good. I suppose if that were to happen a whole growth industry
>> could spring up, much like LibraryThing relies on MARC record
>> information (in addition to Amazon records, or mashups of both sources).
>> I don't think libraries should do it all-- companies like Syndetics and
>> Serial Solutions can do things in a more rational way, but I think that
>> the original mission and purpose of libraries should lead to a set of
>> standards that can form the basis for both library-centered free
>> information access and for a "new economy" model of information services
>> to be used by all of society.
>>
>
> I don't think you are suggesting this, but I think it is dangerous to put
> everything on hold, and not to encourage other approaches, while we get
> the underlying data structures and standards "right", even assuming we
> could agree on what "right" is. There is no guarentee that what we
> consider the most appropriate data structures and standards would spur a
> growth industry. I also wonder if it is wrong minded to divorce the
> availability of the data structures from the availability of the objects
> they seek to serve. There's been a lot of work done on full text indexing
> in particular, and using LSI and other techniques to draw out underlying
> patterns from critical masses of content. Plus, even the most perfect data
> structure can be undermined by the systems that work with it, if the most
> effective form of metadata was targeted, it would probably involve
> discussion of ways of indicating preferred tokenizers and stemmers in
> addition to the semantics of "entities" and "resources" and so on.
>
> I agree that the role of metadata in the discovery layer is definitely
> important, but I really wonder if a blended model of existing metadata,
> outside data, usage patterns, multiple services, and mass digitization is
> a better use of library priorities than seeking the holy grail of data
> structures. On the other hand, if the quest is deemed the priority, and
> maybe it is, it would be worth doing some research on how the technology
> could best leverage such efforts, or could dovetail with their intent. I
> am no expert at what other communities have done with metadata, but my
> sense is that in areas like real estate and pharmaceutical metadata,
> pragmatism has usually won out over every other approach, and these are
> communities that probably have been able to do a fair amount of homework.
> At the very least, I would like to see some more evidence in all this that
> the part of the information retrieval process controlled by metadata can't
> somehow be compensated, even to a small degree, by other parts of the
> ecosystem that the web represents.
>
> art
>
>
--
Jonathan Rochkind
Sr. Programmer/Analyst
The Sheridan Libraries
Johns Hopkins University
410.516.8886
rochkind (at) jhu.edu
Received on Fri May 18 2007 - 11:27:48 EDT