Dear Roger,
Thanks you for clear reply and Richard for his additions.
My problem in understanding your GUID system is that -if I am
right- it aims to integrate two functional properties for names,
namely (1) cross-referencing for nomenclatural mapping and (2)
definite identification of all name objects in the linked systems,
which also function as an input of user-cases.
The Semantic Web Applications for knowledge-based systems I know
always try's to separate those functions. The benefit of splitting is
that the functional constraints on (in our case) name objects will be
restricted depending their use on either the critical
cross-referencing or the usage context. When you merge both functions
you induce conflicting constraint on your name object and I think that
is what happens in your current discussion with Richard: For the
cross-referencing property you need nomenclatural exactness; for the
usage (modelling) context you would like to have a maximum of
flexibility and freedom to treat name objects.
Of course both methods have their pros and cons. A problem of
explicit cross-referencing is that it is an awful lot of work in the
beginning and you need to have certain standards, but once established
it's probably the best system. Merging cross-referencing and
user-cases has the benefit of having a quick start, however,
conflicting constraints on your data (=name) objects, a lot of complex
algorithmic modelling, and always the uncertainty of ending with an
sub-optimal system.
If I am right the GBIF attempt is not 'either-or', but a kind of
hybrid system. Since I am not an expert in this informatics area I
realise that some of the answers probably can be found in recent GBIF
documents on this issues and your wiki page. So time to do some
homework.
Kind regards,
Yde
Hi Yde,
TaxonNames
Yes it is ambiguous as to whether a
nomenclator may wish to issue a TaxonName GUID for 4 and/or 5. What I
would do if I were a nomenclator is issue a TaxonName GUID for both.
The GUID of the wrongly spelled one would return an object which
included the GUID of the correctly spelled one. This is just what I
would do not what a nomenclator may choose to do.
TaxonConcepts
TaxonConcepts depend on intention. From
just a list of words it is usually impossible to say whether they
represent something that should have a TaxonConcept GUID or not.
If this list was entitled "Moths I have caught in my moth
trap" I would argue strongly that you should not treat them
as concepts. They are an attempt by some one to say which taxa they
have found not an attempt to re-define taxa. The authors want to
reference existing taxa not give objects an identity. The items on
the list may get GUIDs from some recording scheme system though.
If the exact same list was entitled "A treatment of Zus
from Far Away Land" then it seems to me that they are all
meant to be concepts (possibly with bad nomenclature). The authors
want to tag/label these taxa so that other people can reference
them when they go bird watching in Far Away Land - or whatever animal
a Zus is...
An
Analogy
I have in front of me a book called
Everyman's Dictionary of First Names. Here are some names
from the book.
- Milli
- Mills
- Milly - see Milli
- Milo
- Milson
- Milton
The question: "Should we issue
National Insurance numbers for these people?" is not a good one.
They are not people they are just names! But we only know that because
I told you where I got them from. The people who compiled the book
probably had a database with IDs on. They could try and set up a
global system for names with GUIDs. This system would clearly be
completely separate from a global system for identifying people by
number for tax purposes although the tax system may refer to the name
system and a credit reference agency might provide a service for
getting an NI number from a person's name plus some other
disambiguation data.
So we can argue about the correct spelling of Milly and the register
can have its own pointers to 'correct' spelling but Milly Smith still
gets a tax bill because when she was born they gave her a number.
(Incidentally I believe Denmark actual has this system. It has a
national list of acceptable names for children and it has a system of
issuing ID numbers to everyone at birth. The UK just gets confused
with NI numbers and NHS numbers etc)
Hope this helps,
All the best,
Roger
[BTW I had the name book to hand because the children are choosing a
name for our new cat. Milo is the favourite but we are open to
suggestions. He is a ginger tom with a sister called Motlie].
Re: GUIDs for Taxon Names and Taxon
Concepts
Dear Roger,
One puzzling thing for me to be explained
in more detail is the following:
Extending the example of
Richard:
1. Aus Smith 1995
2. Xea Jones 2000
3. Aus bus Smith 1995
4. Xea bus (Smith 1995) Jones
2000
5. Xea ba (Smith 1995) Jones
2000
6. Xea bus (Smith 1995) Jones 2000 as it
appears in Pyle 2005
= Xea ba (Smith 1995) Jones
2000
I agree with Richard that it is ambiguous
whether nos 4 & 5 should get TaxonName
GUIDs or TaxonConcept GUIDs, but I
believe this is a matter of defintion we can solve.
However, how to discriminate between the
TaxonConcept of no. 4 and the TaxonConcept of no. 6 which includes
subjective synonymy? I assume you need a GUID for each documentable
usage instance?
Kind regards,
Yde
Hi Rich,
So you define a NameUsage as:
"Any occurrence of a NameString
as it appears or is explicitly implied
within some form of static documentation."
Let us explore this definition! Picking a
volume almost at random (I like the cover) I chose Porley &
Hodgett (2005) Mosses & Liverworts.
http://www.tdwg.hyam.net/images/bryo_01.jpg
and picking a page at random - in this case 136
http://www.tdwg.hyam.net/images/bryo_02.jpg
So we have a 'static' document and it is chuck full of NameStrings. I
have circled some of them.
We have Ditrichum cornubicum (a red data book moss). It is
mentioned, with a picture on the previous page and at the top of this
page it is mentioned a few more times. Further down this page we
have Buxbaumia aphylla which is also mentioned twice. There is
a picture of it on the next page.
So how many name usages do we have here? There seem to be loads.
- Does each mention of the name on the page count as a usage? -
would seem to be a silly thing to do.
- Does mentioning the name on different pages mean different usages?
- would also be silly but we don't have anyway to judge (different
pages within a journal or combined work for example?)
- How about same page but different context? The picture may be of a
different moss to the one that they mention in the text.
- If a subspecies is mentioned does that count as a usage of the
specific name (it has been used) and likewise a binomial implies a
usage of the genus name.
- There are around 1100 species mentioned in this publication. They
are probably mentioned on average 3 times each (a guess) so that is
3300 new name usages. Plus they are all binomials or subspecific names
so double that for the different usages or genera etc. So 6,600 name
usages in this volume. I wonder how many publications like this come
out a year globally?
I really can't see how one would apply
your definition. Perhaps if you restricted it to taxonomic works but
then you have to define a taxonomic work and you are still limited to
how it has to be stated to act as a 'usage'. It certainly isn't clear
to me.
We can easily define what a TaxonConcept is because it implies
intent. If I want to create an object that I want you to refer to
as a definition of a taxon then I am creating a TaxonConcept and
should issue a GUID to make it easy for you to refer to it. If not
then I shouldn't bother. If I want to use the services of a
nomenclator to define the publication and typification of the name I
am using then I can use a TaxonName GUID within my definition - but I
don't have to. I can't see how that can be any simpler than
that.
Porley & Hodgetts (2005) have no intension whatsoever of
'committing' nomenclatural acts or of defining any taxa that people
will later refer to. They are simple referring to existing
concepts. Yet by your definition they have created over 6k name usages
that a diligent publisher might issue GUIDs for.
Have I completely misinterpreted you definition? If so could you
define it a little tighter? If you imply that the author has to have
meant to describe something then you are just creating the
TaxonConcept definition I am working with here. How else can you
subset all the times names appear in print?
This is all great fun but we do need to nail it down and move on.
All the best,
Roger
Hi Roger,
Could you attempt a concise
definition of a
UsageInstance we can all agree on
then :)
Sure: Any occurrence of a NameString as it appears or is explicitly
implied
within some form of static documentation.
"NameString" refers to a string of textual characters meant
to represent a
name of biological organisms. This can be defined more
restrictively to
"ScientificNameString" (names that conform to one of a
designated set of
nomenclatural Codes), or more broadly to include vernacular
NameStrings.
"static documentation" can be defined broadly, to include
publications,
databases, and any other form of documented medium of human
communication.
The "static" part means that it must represent a snapshot in
time. In the
case of dynamic databases, this would require a "date stamp"
for each
UsageInstance -- either for an individual record, or for a snapshot of
the
entire dataset.
The "explicitly implied" part addresses zoological-style
nomenclatural
listings along the lines of what Yde has sent to this list, where a
genus is
listed once as a header, and species epithets are enumerated
below,
explicitly implying a series of binomials, even if they are not
actually
printed on papaer as such.
The point is, the definition is highly flexible, yet mostly
unambiguous
(assuming sufficient metadata for identifying a documentation
instance).
I think it only makes informatic sense to distinguish two "kinds"
of GUID
for taxonomic objects if the distinction between the objects is
unambiguous.
Given these NameStrings:
1. Aus Smith 1995
2. Xea Jones 2000
3. Aus bus Smith 1995
4. Xea bus (Smith 1995) Jones 2000
5. Xea ba (Smith 1995) Jones 2000
It is ambiguous whether there are three, four, or five distinct
NameObjects
represented (i.e., it is ambiguous whether #s 4 & 5 should get
TaxonName
GUIDs, or TaxonConcept GUIDs via SEC instances).
However, given this list:
1. "Aus" as it appears in Smith 1995
2. "Xea" as it appears in Jones 2000
3. "Aus bus" as it appears in Smith 1995
4. "Xea bus" as it appears in Jones 2000
5. "Xea" as it appears in Pyle 2005
6. "Xea ba" as it appears in Pyle 2005
7. "Xea" as it appears in ITIS Nov. 11, 2005 snapshot
dataset
8. "Xea bus" as it appears in ITIS Nov. 11, 2005 snapshot
dataset
9. "Xea ba" as it appears in ITIS Nov. 11, 2005 snapshot
dataset
There is very little ambiguity that each item in this list gets its
own
GUID. Until a universal definition of a "NameObject"
emerges, certain usage
instances can serve as surrogates for basionyms (1, 2, 3), or
botanical new
combinations (4), or TaxonConcepts (1-6) -- in whatever way that a
data
manager needs or wishes to establish linkages among GUIDs (e.g.,
linking #s
4, 6, 8 & 9 to #3 via "is basionym of"; or linking #s 4,
6, 8 & 9 to #2 via
"is combined with"; or linking #5 to #6, #2 to #4, and #1 to
#3 via
"contains").
Several people have expressed a desire for "simple" and
"flexible", and I
think this approach maximizes both.
Aloha,
Rich