At 21:46 19/12/96 +0000, Andrzej wrote:
>Representation methods *do* make a difference, but mainly affect how
>efficiently you can manipulate information.
I agree, but I was trying to move to discuss the problems at the foundation
level - once the foundations are right, different representations may be
used where they are optimal.
>>Once we have defined (if it is possible) these standard concepts
>Absolutely right! The hard part is finding the most effective
>method.....but are there universal "standard concepts"? There is a lot in
>the literature to suggest that conceptualisation is task and purpose
>specific.
I also have very strong reservations about task and context independent
concepts, (which was why I added the 'if it is possible' rider) - but this
applies whatever the representation. Just because you use SGML doesn't solve
this fundamental problem.
>>So can I make yet another plea that we drop discussion of representation as
>>irrelevant, and focus on the concepts and their semantics? This is the
>>interesting area.
>
>"Representation" has drifted in meaning, I think. It now commonly includes
>the "contents" (semantic models, constraints, ontologies, etc., and
>extensional knowledge) rather than just the means. But I think there is an
>additional component:-
I specifically meant representation in terms of SGML vs other strategies.
Not the more general meaning of 'representation' in AI. Mind you, here is a
good example of 'concept drift' - however well you define something it may
mean something else by common consensus next year.
The purpose of my message was to say - can we focus on the semantics rather
than the particular implementation, which as you have pointed out, may not
be in vogue next year.
> - what are you going to use the information for, and how do you know what
>you need?
Exactly. This to me is one of the achilles heels of EMR systems. Doctors
tend to enter data in EMR's like they did paper notes, which really means
they have a single main 'task' in mind - to remind them what they were
thinking at the last consultation. There is another task usually present
which influences note taking - the 'what if I get sued' task. It is
difficult to use paper records for more than these two tasks, because
searching them is so difficult. However, EMR's open up many more possible
tasks - Epidemiology etc. However, existing terminology systems (GALEN
included) and EMR systems have not been designed to be task independent.
What is worse is the Doctor hasn't been designed to be task independent in
the way she/he records the consultation. Consequently, the use of the EMR
for purpose other than the two tasks that were in the mind of the a) doctor
using the system b) terminology designer c) EMR designer, will fail. And
have failed, repeatedly. When clinical trials are undertaken, specific
training in data recording for this new 'task' is mandatory, with carefully
defined concepts.
My area of interest in this is to aid decision support systems - this means
using the EMR for a task which the user did not have in mind (most likely)
when he recorded the entry. The reason why I still pursue this goal, is that
there are workers (only a few, I admit) in AI who believe that it is
possible to get to the holy grail of task independent knowledge, and a great
many who believe it is possible to get close to this, if not actually reach
it. What we should be able to achieve is a great improvement on the current
situation. That may be enough to achieve what we want. But I'm sure the use
of SGML per se isn't going to solve these problems.
Pete
---
Peter Johnson
[log in to unmask]
(+44) 1525 261432
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|