> However, what constitutes 'the same thing' depends on the need at hand. This
> means requesting from a FRBRized display is inherently problematic --
> particularly if the patron thinks s/he is requesting something specific in
> mind, and doesn't realize that the requesting algorithm simply chooses the
> most popular work, the one that's easiest to get, etc.
I think the great problem with FRBR is that it's binary. Pre-digital
library systems were all binary, mostly because of physical
limitations. So, something either is or isn't on the shelf, or part of
the subject "Love Stories." Relationships between books aren't binary.
Sometimes it simplifies things to make them binary—but I think most of
the benefit is for the catalogers mind, not the users. Users can
understand that books relate to each other in complicated, shaded
ways.
Case in point. LibraryThing's "Combiners!" group is convulsed by
debate about LT's work system—a drawn out battle between lumpers and
splitters, and between people who like my social conception and those
who want work-set relationships exclusively based on content
differences. The truth is, there is no truth. There's just what is
more useful and what is least, and that depends on the data and the
searcher. I know we need a better system—a better system, and other
systems.
Making a stab at how items relate to each other is great. But I fear
that FRBR is ultimately another totalizing attempt to shoehorn reality
into boxes. This will be a real problem as algorithms start to
approach the same issue. I'd bet you that Google could already come up
with a pretty nuanced, non-binary map of relationships between, say,
all the editions of Whitman's poetry based on full text. Wouldn't that
be fascinating? It might have some guesses, but the overall result
would be richer than a FRBR model could represent.
So, I submit that if the library world fixies on another standard
rooted in the binary limitations of the physical world, it will miss
the boat again.
Tim
Received on Tue Feb 17 2009 - 14:27:51 EST