Dear Roger,

Many thanks for developing the taxon concept wiki. You might want to take a look at Nadia Anwa's work at the Univ. of Glasgow. She has developed a taxon name warehouse that attempts to automate the mapping of related taxonomic names to build potential taxon concepts (http://spira.zoology.gla.ac.uk/ - be patient with the server, it is a little slow). Each search generates a map showing these relationships. In this scenario GUID's would be associated with the component names and the map (or rather the potential taxon concept derived from the map). Nadia's approach is perhaps less refined than what some people envisage but is a first step toward automation.

Regards,

Vince

____________________________________________________

Dr. Vincent S. Smith Illinois Natural History Survey 607 East Peabody Drive Champaign, IL 61820-6970 USA Tel: +1 (217) 265-0831 Fax: +1 (217) 333-4949

E-mail: vsmith@inhs.uiuc.edu Web: http://darwin.zoology.gla.ac.uk/~vsmith/ iChat Video Conferencing: vsmithuk@mac.com (invitation only) Skype: vsmithuk; or try SkypeIn (call whichever is nearer you) SkypeIn London: 020-755-88950; or Chicago: (312)-239-0950 ____________________________________________________

On 11 Nov 2005, at 07:26, Roger Hyam wrote:

Hi Rich,

I think you are wrong in your conclusions that we do not need GUIDs for TaxonNames and TaxonObjects but I need to do diagrams to illustrate the case so have created a wiki page here:

http://wiki.gbif.org/guidwiki/wikka.php?wakka=SeparateNamesAndConcepts

I also intersperse some comments below.

Richard Pyle wrote:

...

Hi Roger,

...

Rich - yes. I think you sum up the difference between new combinations in ICZN and ICBN well. But... just because the ICZN does not consider the usage of a name in a different genus as a nomenclatural act does not stop us creating a data object (TaxonName object) to represent what that name looks like in the new genus - perhaps with its new ending and possibly different author string (ICZN Recommendation 51G).

I agree that nothing is stopping us, but I forsee headaches down the road as a result of doing so. In your next message, you wrote:

"1. We have two kinds of GUID (one for TaxonNames and one for TaxonConcepts)."

...

From this I interpret that TaxonName GUIDs are different with

TaxonConcept GUIDs -- is that correct? If so, the problem is that a botanist would assign a new TaxonName GUID to a new combination, and a zoologist would assign a TaxonConcept GUID to the same entity, because "genus combination" is a property of a name in botany, and a properyy of a usage (~concept) in zoology.

Yes, you could certainly force-treat zoological names as though they were botanical names (treating new combinations as "new names"), just as you could easily force-treat botanical names as though they were zoological names (assigning TaxonName GUIDs only to basionyms, and representing new combincations via Usage/TaxonConcept GUIDs). I just believe that we will come to regret it if we leave the distinction "fuzzy".

I am not suggesting we leave it fuzzy I am suggesting we leave it up to the nomenclators and that it isn't a GUID issue. I don't leave the maintenance of the brakes on my car fuzzy by not fixing them myself - I delegate it to a mechanic. Just because we are not solving the problem here does not mean that we are not going to get the problem solved.

...

My point has always been, and continues to be, that *IF* you have separate "kinds of GUID" for TaxonNames and TaxonConcepts, the line between the two should be unambigious (and ideally be consistent for both botany and zoology). After thinking about it some more, in the context of what has been written on this thread, I find myself coming back to that first "IF". Given that there seems to be a need and a desire to leave the definition of a "name unit" (to which a GUID is assigned) loose and flexible, then perhaps it would be premature to establish a GUID system for Name-Objects at all. Instead, I think we can both simplify *and* disambiguate taxonomic objects if there is only *ONE* GUID system -- which represents a Name-Usage instance.

Then don't refer to the TaxonName GUIDs issued by nomenclators when you issue your TaxonConcepts. Just ignore the nomenclator bit. If you are correct then everyone will follow your lead. A two GUID system will just degrade to a one GUID system if the names thing is wrong.

...

In the case of datasets consisting of only namestrings, with no specific implied concept objects, the names can be interpreted as "NameString SEC Nobody" (=Nominal TaxonConcept in TCS). Nomenclators could use whichever subset of these Name-usage GUIDs that they wish to refer to their own version of a name-object. For example, ZooBank can concern itself only with those GUIDs attached to original basionym usage instances, and IPNI can manage both basionyms usage instances and new-combination usage instances. ConceptBank could expand the scope of GUIDs to all those usage instances that represent defined concepts. Name indexers could use the broadest set of GUIDs (effectively all name-usage instances).

I think you are mixing issues here. The name string probably represents a TaxonConcept sec the dataset - depending on your intent. If you have had people scoring observations to a checklist for an area for a number of years then it would be convenient to treat the names as concepts so that you can reason about how they may be related to other better defined concepts. They are concepts sec your list. If they are just words you found in some books then they are just words till you intend to do something with them.

A nomenclator should not be publishing data other than nomenclatural data (debatable what this is I know but no circumscription data anyway).

ConceptBank sounds like any indexer (like GBIF) that crawls the people who are defining concepts of taxa and tries to provide sensible query expansion on the basis of the GUIDs or however else it links the objects together.

...

So, in summary, my feeling is that if we are to think of "two kinds of GUID" for taxon objects, then we need an unambiguous (and cross-Code) distinction between the two. If such a distinction is too cumbersome to draw (as it seems to be, based on the current thread), then we should only go with one kind of GUID -- and by default it should be the one kind that is most flexible (=usage instance).

See the wiki page for why I think this is an erroneous conclusion.

Here is a definition of the two types of GUID: A TaxonName GUID resolves to a data object that only contains information about nomenclature. The provider does not intend anyone to be able to identify a specimen to this GUID. TaxonConcept GUIDs resolves to a data object that contains information about the delimitation/circumscription/relationships of a taxon. The provider intends people to be able to identify or otherwise related data to this GUID. I think the notion of the intension of the provider is very important as it answers many questions. "Should I attach a GUID to object X?" depends on what you want other people to do with it. I don't think this is very cumbersome definition? There is no 'natural' distinction between these two things. It is a pragmatic split so we produce a series of 'rooted' directed graphs of

All the best,

Roger

--

---

Roger Hyam Technical Architect Taxonomic Databases Working Group

---

http://www.tdwg.org roger@tdwg.org

+44 1578 722782