Hi again,
I just had a shower (on my way to bed), but had a thought;
On 6/21/06, Bernhard Eversberg <ev_at_biblio.tu-bs.de> wrote:
> Are these indexes really browsable, up and down (and displaying
> not just lines that somehow match the user's input but beyond that
> as far as they want to go?)?
We use the indexes to cluster information, so in that sense, you can
browse around in as many clusters as you possibly can want (meaning
more spaces than there are indexes). If you're referring to subject
browsing, no we don't really do that because it hasn't appeared that
we need to; clustering seems to do that for us in the sense that the
semantics of the search you just performed is right there on screen.
Although, your subjects and titles and whatsnot is there as clusters,
and you're free to explore these if you like. I have another
application where you can browse around in a big thesaurus and see
what items it finds at each term. It works really well, so do we need
any other way? Let's find out through testing.
> All of that, besides being tied in with one language, will not do for
> personal or corporate names.
Of course it does; statistical error pruning is exactly what Google
does to give suggestions. But more importantly; what is the definition
of what "will do"? We certainly don't design librarian systems; that's
how most ILS and OPACS are right now, with fielded searching and
knowledge of LCSH and Dewey and what you'll find in a MARC 245
subfield c and so forth. We're dealing with "most people" here, and
for those I think statistical error pruning will do just dandy. But if
you have another important user-group to which this "won't do", then
do tell.
> What any plain decent good enough catalog
> should display
Should do or actually do?
> is something like this (the arrows pointing to the
> authority spelling, where it has been linked, in this case different
> from LC's):
Hmm, yes, authority records are well and fine as long as they are
maintained, but I challenge you to find any books that I've been
involved in using these records using any of my spelling varieties
(there aren't that many though :) ; I'm too small and insignificant
and too new to even have such a thing. There needs to be a balance
here, somewhere between doing nothing and doing all, for those who are
truly known and those who are in the margins. We're trying very hard
to reach some middle ground, which seems to be still far above the
average OPAC of today.
> 17 shostrom, everett l
> 2 shotam, nirmala puru -> puru shotam, nirmala
> 1 shotbolt, charles r
> 1 shoter, l
I'll add one for you ; sjostakovitsj, norwegian spelling. There's only
so much that makes sense to put into these files, and I seriously
believe that these things need to go in the "analysis" folder more
than hand-spun metadata. But that's just me; there's others out there
with different answers.
> If this looks too exotic, try find Stephen Hawking's works in
> LibraryThing. [clue: try Hawkin, Hawkins, Hawkings as well]
This is the same reason people have trouble with tagging and other
means of classification. It's nothing new, nor is it easily solved,
not even with authority records. I think we need to be a bit pragmatic
with these things, and we experiement with letting clusters tell you
more about what you've found. Maybe, just maybe, the idea of authority
records could be made redundant through clever design. Who knows?
Now I'm certainly not very fond of second-guessing what is "good
enough", especially not for our users. Test it out and see if it is
true instead. If rigid authority record use constitutes a want in less
than 1% of users, is it still something we should pursue?
I'm all for testing these things instead of saying "any good system
must do X, Y and Z"; how else could you know?
Alex
--
"Ultimately, all things are known because you want to believe you know."
- Frank Herbert
__ http://shelter.nu/ __________________________________________________
Received on Wed Jun 21 2006 - 08:58:58 EDT