On 8/8/07, Rinne, Nathan (ESC) <RinneN_at_district279.org> wrote:
> OK, my last post for the day - and I think for at least 3 weeks. (sorry
> Alexander, if you were actually looking forward to a response :) ).
Yes, I am disappointed ; your questions / challenges are the epitome
of the current library worlds geist, and it's a bleeding shame that a
lot of it is just left hanging like this. Having said that, there's
been a lot of other really great responses too.
Now, Will has given some more clarification of the AI thing, but I'll
add a little bit too it since it seems that there's a technological
divide and confusion about technological possibilities within the
library world ;
AI is mostly about developers trying to make software / hardware make
human decisions about data. It's not even that hard to do for a lot of
things, and its applicability (especially the fuzzy logic) you'll find
in plain appliances like cameras, cars, security equipment, etc. And
it's not really hard, works fantastic, and you don't even notice that
they're there (which is a very subtle but important point!).
No, the problem with AI is that it requires a) people that know about
this stuff, b) support, c) time, and d) patience. At the moment the
library world has *neither* of these things when this problem-space is
discussed. How can I explain what AI can do for us, when I'm told to
make a prototype? How can I make a prototype if not given the time,
don't have the meta data, nor anyone to help me out? Just me hacking
this stuff in my spare time is just not going to cut it.
So instead of trying AI, we do what we always do ; chuck an index
across some data fields, boolean logic in the search query, and
hopefully someone finds it useful. Bah! People have been doing this
for years and years, but it's just not good enough. We replace one
such system for another, hoping that the few extra features are going
to cut the mustard for us.
Google and others are working very hard at the next step, which is
more like latent semantic analysis, flock / life patterns in infosets,
fuzzy logic across indeces, all based on deterministic models of human
behavior, language and "aboutness." Where are we in this venture? I'll
tell you where we are ; under a rock.
As to RDF, it's just a metadata model and formats with a very clunky
ontology language on top. Topic Maps is a slight improvement (better
formats, rock solid PI system, and a slick data model), but not as
agile / fragmented. These things just ship metadata back and forth,
and has got nothing to do with AI per se.
I used to do real-time image pattern recognition, and humanistic data
value models. And now I spend my time making interfaces over Lucene
and bickering on this mailing-list. Go figure. :)
[...]
> Maybe I am wrong, but I don't have the confidence that other folks on
> this list really understand my concerns. I would love to be proven
> wrong though, although I am not even sure at this point where to start
> in such a discussion.
Well, I think you're arguing contextual completeness, but you're never
going to fit that into the current cataloging scheme. For us all to
reach goals which I actually think we all got (better systems, better
meta data, better service, better sustainability, better libraries,
etc.) we need to take some brave steps away from what we're currently
doing, whether Mann agrees or not.
> Having just read Eric Lease Morgan's last post however, I am not even
> really sure people even think that discussion is really worth having.
> Let these treasures of the past go... let the brave new world come. In
> my mind, technology is loosing its proper place (I never thought I or
> Mann was anti-tech! until...), and we have become drunk with it, to go
> along with our being obsessed with the "instrumentalization" of all
> knowledge (which after all, will also *soon* be able to be downloaded
> unto a computer) :)
You're missing the point, and I'll reiterate that the "treasures of
the past" are highly valuable, it's the very things that will save our
profession, but unless we can harvest it and use it in better computer
systems, we're dead. You tell me ; how can we harvest librarians
knowledge of LCSH? How can we present that value in a search system?
What secrets of Dewey can we harvest for better knowledge management?
Our amazing knowledge of "how to find stuff" needs to be exploited by
systems that can handle the amount of information, but how?
> Done fighting... for now. :)
Chicken. :)
Regards,
Alex
--
---------------------------------------------------------------------------
Project Wrangler, SOA, Information Alchemist, UX, RESTafarian, Topic Maps
------------------------------------------ http://shelter.nu/blog/ --------
Received on Tue Aug 07 2007 - 18:50:45 EDT