Some scream-of-conscisousness (c) thoughts about med diagnosis E/S HOW-TO.

Hello all.  I haven't touched an E/S or PROLOG in a looong time so I
don't know what's popular these days. briefly scanning COMP.AI I see
references to OPS5.. SO.. what ever happened to SNOBOL??

but, seriously... I'm surprised to note that folks still think you
need to translate human thought into some kind of linked-list using a
made-to-order "language" like PROLOG, OPS5 or whatever.  far as I can
see, medical diagnosis E/S haven't really gone where we'd Hoped they'd
go since someone invented the terms...

Decision-trees, neural-nets, linked-lists (which is really what you
get out of any E/S 'programming language').. fine but maintainence is
a umm... is difficult.  Why not go KISS?

"Knowledge" just needs to be -written down-,made accessible and
understandable, e.g. textbook, that's how we're taught to learn.
Finding the answer is just a matter of remembering where to look...or
having a tool to assist that endeavor.

So, 2003, we've got the world's biggest data/info/knowledge base you
ever DID see right Here.  Well GOOGLE's 3 billion indexed webpages is
huge no matter HOW you slice it.  It's just not a Medical E/S.  But it
Could be.

Ya got all the elements you need for an E/S just SITTIN' there..

   massive storage (free) on multiple servers 
      [YAHOO, LYCOS, et al]

   unrestricted content: 
     encode to your hearts content, use good keywords tho
     and hey you can put ENGLISH (or any language) 
     text on the page....hmmmm

   some unconCEIVably fast access software (well 
    compared to pre-WWW days)
     aka search-engines / indexing tools 
       [GOOGLE, ALTAVISTA, DOGPILE...]

so you just tell me how doing a GOOGLE query for 
  VW SQUAREBACK 1972 ILLINOIS
is so different from encoding symptoms and then having a specialized
system go hunting (depth? or do you prefer breadth?) for possible
matches to use in
subsequent (inner) searches?

other than taxonomy and a little finagling, I don't SEE any
difference.
except my VW query isn't contextually understood or restricted ONLY to
VW sites or ILLINOIS sites, it's more a case of BFMI**, to borrow a
famous acronym.
you always get waaay more than you need.

in medical diagnosis that's OK.. the sin of omission will 
kill More people Faster than its counterpart.

I guess my question, then, is:  Anyone persuing what you might 
think of as a back-door [non-Prolog, non-OPS5, non-whatever] 
approach to medical diagnostic DSS or E/S utilizing WEB technologies
to do ..well.. preCISELY what they do BEST.. FIND stuff that's
keyword (symptom..) related.

N.B.: a 'relation' ['hit'] in this case is any searchterm occurring on
more than one page, excluding 'noise words'.

and yes I hear the fuzzy-logic and Bayesian folks going But how about
the tweaks and probabilities?

well, hey, spend more time getting good history and symptom work-ups
and use that info to drive the initial search. apply inference and
fuzziness  further down the line, maybe?  no sins of Omission,
remember?

My apologies, I'm just thinking out loud :)

regards,

terry b, chicago, Illinois


** BFMI...brute force, massive ignorance