At 08:10 AM 7/20/00 +1000, Kevin Thiele wrote:
At 11:56 PM 18/7/00 -0300, Bob Allkin
wrote:
[...]
1) description of
distribution (a hierarchical descriptor - with attributes
attached to each substate such as native/introduced/etc)
2) description of use [e.g. taxon- plantpart(eg leaf) - forwhom (eg
children) - forwhat (eg sorethroat) - how (eg infusion)]
3) description of ecology - requiring links between two
"descriptors" (eg.
plant A is a tree at altitude X BUT plant A is a bush at altitude
Y)
then Im less clear how these would be accomodated. Is
there elsewhere a
set of descriptor/state structures that I could see?
[...]
Can you suggest a way to accommodate your suggestions?
The examples Bob gives are very complicated (higher order logic? [he
suggests quickly getting in over his head in formalisms!]). At
least part of the complication derives from the fact the
"subject" is a species (or taxon) rather than a specimen, and
is necessarily a summary of primary data.
I think we talked earlier about the simplest sorts of statements taking
the form of a logical triple:
[THING] has [ATTRIBUTE
CLASS] of [ATTRIBUTE VALUE]
[fish-123] --> [dorsal
spines] --> [11]
A simple (minded?) solution to Bob's example #3 might be to
accommodate the complexity by narrowing the scope of the THING:
[plant A at altitude X] --> [growth form] -->
[tree]
[plant A at altitude Y] --> [growth form] --> [bush]
This might be a bad approach... whatever. The point I
want to make is that the statements above are simple, whereas statements
like
PART-A of TAXON-B is used by CULTURE-C for FUNCTION-D in
SITUATION-E
are so close to natural language, that the software/standards
problem is WAY beyond our capabilities. I think we should designate
this level of complexity as "out of scope".
I would rather see us focus on the scope implied by the union of: 1)
basic description (Delta), 2) identification (IntKey, Lucid, etc.), and
3) phylogenetics (PAUP and McClade). I think that's a pretty
significant challenge.
-- finally --
Would it be possible to get "data models" (or at least
simplified "core" models) of the relevant applications posted
or pointed to? I think we're operating with a severe handicap until
we have them.
-Stan