interactive keys and others
M.J.Dallwitz at NETSPEED.COM.AU
Mon Mar 13 17:19:11 CST 2006
Mary Barkworth wrote:
> One quickly becomes aware of how unparallel many of our descriptions are.
> This is not completely bad - a key is supposed to be a shortcut to
> identification, whatever its format. A description, if we separate them from
> keys for a moment, should be permitted to contain more than is in the key -
> if more is known. Otherwise we find ourselves limiting what is in the
> description. There should be similar core information in all descriptions.
> It is also helpful if there is a level of uniformity in phrasing throughout
> - but again, one may wish to enhance the information about some aspects for
> taxa that one knows particularly well. If we require absolute parallelism,
> we lose the additional information that is available, something that is not
> helpful. Yes, it would be lovely if one could then go out and obtain the
> additional information for all other species; that is not always feasible.
> Should we conceal what is known in order to keep parallelism?
By 'parallel', I assume you mean descriptions in which corresponding parts
exist in all the descriptions, e.g.
This is neither necessary nor sufficient in data intended for use in
identification (or classification). The requirement is for _comparative_
descriptions, i.e. those based on properly formed characters. The list of terms
is not a character, because the 'states' are not mutually exclusive. I call
it a pseudocharacter, because its structure makes it look like a character,
and it would be accepted as a character by any program for processing
descriptive data. (I realise that there are other uses of the term.)
I completely agree with what you say about not limiting what is in the
description. In DELTA format, non-comparative information is easily
incorporated in the form of text 'characters' and comments, without
compromising the data for comparative purposes. For example, a character
could be defined as
and the above data recorded as
petals red (scarlet)
It's certainly not necessary, or even advisable, to record every character
for every taxon (provided that your software copes properly with this). The
effort required to do that would be better spent in adding data where they
are most needed. The 'Diagnose' option in Intkey gives guidance in this. It
finds sets of taxa that are not separated, or are not separated with
sufficient redundancy to provide alternative characters for identification.
This is important so that users have a choice of characters, and for the
proper operation of error tolerance. When these areas of deficiency have
been found, they can be improved by recording relevant missing data (if any)
in existing characters, or by adding extra characters to separate the
problem taxa, and recording them for those taxa.
Contact information: http://delta-intkey.com/contact/dallwitz.htm
DELTA home page: http://delta-intkey.com
More information about the Taxacom