W3C home > Mailing lists > Public > public-lod@w3.org > December 2009

Re: Species Concept Mapping RDF fixes and question, should the species be represented as a class? Class SpeciesConcept => Class Species Cougar

From: Bob Morris <morris.bob@gmail.com>
Date: Fri, 4 Dec 2009 12:32:19 -0500
Message-ID: <16957b040912040932t6d3a1b5clf90ad42e0b5221a0@mail.gmail.com>
To: Peter DeVries <pete.devries@gmail.com>
Cc: public-lod@w3.org, dmozzherin@gmail.com
On Tue, Dec 1, 2009 at 10:14 PM, Peter DeVries <pete.devries@gmail.com> wrote:
> Hi Bob,
> I came about this after it was suggested to me that a species might be best
> represented as a class.


> It also occurred to me that an ontology that works to describe mosquito
> species would probably be very different than an ontology used to describe
> members of the cat family.

Ah, perhaps if you want to get into descriptive data, yes, but that's
a whole different set of use cases from the ones usually supported by
occurrence data, although sometimes people use occurrence data as a
proxy for abundance, and weight characters and/or states by abundance.

In the end, for ecological observations, it is likely that it will be
necessary to be able to talk about several species at once; e.g. when
it is important to determine if there were two particular species
observed at the same time and place. To me, this alone suggests that
important applications will need a merged ontology at least as to the
species (concepts).  There is fair concensus that the number of known
species is roughly 1.8M---so concepts perhaps 2-3 times that (???).
Also, it is well and often pointed out that for even simple reasoning,
one may have to align the species concepts, not the names, to decide
whether occurrences reported in different data sets are really talking
about the same kind of thingie---which is why you and TDWG model
concepts.  So from that I deduce that for interests of people who
study communities and their interactions, rather than those who study
taxonomic groups, making a species (concept) be a class instead of (or
in addition to)  an instance will lead to as many as several millions
of classes (or several hundreds of million if we find the other 28M
species before we destroy them all....). So I guess I believe that
modeling species (concepts) as classes is not scalable.

A less alarming estimate might be made by asking something like this:
given a survey project, e.g. the NatureServe NaturalHeritage databases
that requires an integrated ontology able to at least to describe
occurrences of the all the taxa that the project describes, roughly
how many taxon concepts are required in the ontology? .  This removes
the burden of having an agreed upon "All taxa ontology" but raises the
difficulty of ontology matching across projects. (That is a studied
problem, though, e.g. in the thesis work of David Thau).



> What I thought was that there are times when you want to treat a species as
> an instance and other time you want to treat it as a class.

See my posting in response to Pat's remarks on the subject,

> Thinking out loud here.
> The lightweight representations shown in this
> example http://rdf.taxonconcept.org/ses/v6n7p.rdf
> are used when you just need something simple that gives you the basic
> information and maps concepts.

Simple is good for data discovery and integration.

> However, it really does not document what you mean by the URI. By that I
> mean, it does not provide any information that will allow you to
> determine what species concept you should apply for a given specimen.

Ah, well, my position, especially as one of the TDWG SDD architects,
would be that how to make this determination is fundamentally a
question of descriptive data in the sense of the nomenclatural codes,
at least for concepts that correspond to a valid name or synonym of a
valid taxonomic name.  This raises really interesting reasoning
issues, because people place specimens in concepts all the time based
on descriptions that are not the published one, Rather, they are based
on other knowledge, especially geographically and temporally local
knowledge.


> For that you will need something more complex, that can be loaded as needed.
> Something more like this representation.
> http://rdf.taxonconcept.org/owlses/v6n7p/2009-12-01.owl (Initially an
> individual file made with Protege)
>
> How would these be used?
> Lets say that an individual Cougar was observed within the boundaries of the
> state of Wisconsin.
> That individual is an instance, however, from that you should be able to
> reason that the species
> http://rdf.taxonconcept.org/ses/v6n7p
>
> was observed within the State of Wisconsin.

If by reasoning you mean formal logic, and if you want to stay in OWL
DL or some other computationally tractable RDF based language,  I am
inclined to guess that you will be forced to make this conclusion
somewhat tautogical. In general you won't be able to use a single
predicate, e.g. isObservedIn to act on both instances and classes
because the subject of a triple is always an instance of something,
and OWL DL prohibits anything from being both an instance and a class.
 So you are likely to end up with an axiom---a definition if you will

C isSpeciesObservedIn W =(def)  thereExists o an instance of C AND o
isIndividualObservedIn W

Unless I am wrong about about OWL2 Metamodeling as in my response to
Pat, I suspect you will be forced to the above kind of solution


Less formally, the observation of a species in a place, will by
definition be the observation of a member of that species in that
place. This seems to coincide with common understanding also.


>From this it will indeed be easy to reason what you want, but it
doesn't seem to deep to me since it will just be a definition.



> This should allow you to then run queries on species observed in a
> particular geographic location as shown in these examples.
> http://about.geospecies.org/sparql.xhtml
>
> You should also be able to make other assertions at the species concept
> level based on data from collections of individuals.
> For example, the mosquito Culex territans femalesfeedPrimarily on Anurans
> (frogs and toads)
> How do I determine if the cat I have captured is Puma concolor vs Puma
> yagouaroundi? (or another less obvious example)
> See
> http://rdf.taxonconcept.org/ses/v6n7p and  http://rdf.taxonconcept.org/ses/Hq5OE
> and from there get the related owl documents. The owl documents should
> provide some information that will allow you to determine
> which concept is the best match for the captured specimen.
> At least that is what I would like to do. :-)
> - Pete
> On Tue, Dec 1, 2009 at 8:30 PM, Bob Morris <morris.bob@gmail.com> wrote:
>>
>> On Tue, Dec 1, 2009 at 8:40 PM, Peter DeVries <pete.devries@gmail.com>
>> wrote:
>> > Hi LOD'ers,
>> >[...]
>> > I was thinking that the species itself should be a class so that
>> > individuals
>> > of that species would be instances of that class.
>> > Probably another skos:Concept class.
>> > So an individual species concept class like that for the Cougar would be
>> > an
>> > instance of a skos:Concept (SpeciesConcept) class and also be a
>> > skos:Concept
>> > class (Cougar) of it's own.
>> > Individual animals would be instances of the skos:Concept class
>> > (Cougar).
>> >[...]
>>
>> Umm, if every species concept is a class, about how many classes, in
>> your estimate,  would there be in a comprehensive ontology?
>>
>> --Bob
>>
>>
>>
>> --
>> Robert A. Morris
>> Professor of Computer Science (nominally retired)
>> UMASS-Boston
>> 100 Morrissey Blvd
>> Boston, MA 02125-3390
>> Associate, Harvard University Herbaria
>> email: ram@cs.umb.edu
>> web: http://bdei.cs.umb.edu/
>> web: http://etaxonomy.org/FilteredPush
>> http://www.cs.umb.edu/~ram
>> phone (+1)617 287 6466
>
>
>
> --
> ----------------------------------------------------------------
> Pete DeVries
> Department of Entomology
> University of Wisconsin - Madison
> 445 Russell Laboratories
> 1630 Linden Drive
> Madison, WI 53706
> GeoSpecies Knowledge Base
> About the GeoSpecies Knowledge Base
> ------------------------------------------------------------
>



-- 
Robert A. Morris
Professor of Computer Science (nominally retired)
UMASS-Boston
100 Morrissey Blvd
Boston, MA 02125-3390
Associate, Harvard University Herbaria
email: ram@cs.umb.edu
web: http://bdei.cs.umb.edu/
web: http://etaxonomy.org/FilteredPush
http://www.cs.umb.edu/~ram
phone (+1)617 287 6466
Received on Friday, 4 December 2009 17:32:48 UTC

This archive was generated by hypermail 2.3.1 : Sunday, 31 March 2013 14:24:24 UTC