Dear Roger,
I
could not reall yunderstand what you and Kevin mean by Characters and
States.
From
the example you gave, it appears to be like classes
(character) and attributes (states) or you want to assign GUIDs to
something like Thesaurus entries? But maybe I completely
misunderstood what you meant..
best
regards,Robert
Kevin mentioning Characters and States and GUIDs
got me thinking and I was wondering if we could cover something along these
lines before the meeting. Please excuse me if this has been dealt with on the
list. I will use a Delta type illustration to my point. This may not apply to
SDD so much - apologies if it doesn't but I am trying to get at a general
point. My comments may be more general to GUIDs though...
When we are
dealing with GUIDs we are talking in an Open World model as opposed to a
Closed World model. If I search Google (open world) and don't find something
it isn't because it doesn't exist - it may exist but not be found for a host
of reasons. If I search my local SQL DB (closed world) and I don't find
something then I can safely assume it isn't there. (This may be a naive
description of Open vs Close worlds but it illustrates the
point).
Taking this to the Characters/States model. We have a character
that looks like this:
Flower Colour (GUID_c01)
- red (GUID_s01)
-
white (GUID_s02)
- yellow (GUID_s03)
And I score a taxon as
"Rose has flower colour red". If I have given GUIDs to the states then
I don't need to use the GUID for the character. "Rose has s01" is fine as the
character is implied.
Can we assume from this statement that my rose
does not have white or yellow flowers? Yes - but only if it is a closed world
and we know that the character never changes (or hasn't changed since the date
of the assertion). If the choice when scoring had been:
Flower Colour
(GUID_c01)
- red (GUID_s01)
- white (GUID_s02)
- yellow
(GUID_s03)
- dark pink (GUID_s99)
I may have chosen "Rose has s99"
of "Rose has s99 and s01" but I simply didn't have that choice before.
So the thing that is troubling me is that Character/State uses a
closed world model where not finding something implies that it doesn't have
that attribute. In an open world system one can only draw conclusions from
presence not absence. We could give GUIDs to characters and states but it
doesn't get us very far as it doesn't permit us to re-use or extend them in a
simple way. (sure you could build an inheritance model for characters and
states but this rapidly becomes a complete ontology language of which there
are a few already available!).
My gut feeling is that in the long term
the Character/State model doesn't transfer well into an open world model. I
suspect this problem may occur in other descriptive areas where the existing
model specifies noun-adjective pairs that I don't have experience of. Perhaps
we could explore this a little. Perhaps my guts need straightening
out!
Your thoughts greatly appreciated.
Roger
--
-------------------------------------
Roger Hyam
Technical Architect
Taxonomic Databases Working Group
-------------------------------------
http://www.tdwg.org
roger@tdwg.org
+44 1578 722782
-------------------------------------