[tdwg-content] Implementing Darwin Core in RDF

Steve Baskauf steve.baskauf at vanderbilt.edu
Tue Aug 16 20:32:06 CEST 2016

I was the lead author on the DwC RDF Guide, so I can try to answer your 
questions about it.  The TDWG RDF Task Group is still in operation, 
although it hasn't been very active for the past several years.  The RDG 
TG has an online "home" at the TDWG Github site.[1]  However, the 
content didn't survive the migration from Google Code very well, so it 
takes some effort at this point to sort through it.  The TG also has an 
email list [2] but there has been little traffic on it recently.

*Dereferencing of the DwC IRI namespace* - Unfortunately, the dwciri: 
namespace terms don't dereference at the present time.  This needs to be 
corrected.  I've created a Turtle serialization [3] of how I think the 
RDF should be written for the dwciri: terms, but it isn't served when 
one attempts to dereference the terms and hasn't been incorporated into 
the official DwC repository.  Part of the problem here is that the 
guidelines for documenting terms in machine-readable form are still 
going through the adoption process.[4] I'm hopeful that when the 
Documentation Specification is ratified, we can make sure that all 
existing DwC terms dereference in a consistent manner.

*Best practice for connecting containers together* - By this, I'm 
assuming you mean linking instances of the various Darwin Core classes, 
or in RDF terms, linking nodes.  The RDF Guide is silent on how to do 
this.  That's not great from the standpoint of actually turning Darwin 
Core records into RDF, but it was a way to complete the guide in a 
finite amount of time.  What is missing is a consensus domain model that 
would lay out how instances of the Darwin Core classes would be linked.  
Such a model should be developed, but that has not yet happened.  Again, 
there is a draft standard submitted for review [5], which if adopted 
will specify (in Section 4) a process for developing such a model.  When 
we wrote the RDF Guide, we provided ancillary documents [6], which 
included examples that followed the RDF Guide and linked instances using 
various proposed models.  There are links to web pages containing 
examples using TaxonConcept, BiSciCol, and Darwin-SW object properties 
to link class instances.  I am not sure whether there is any RDF "in the 
wild" for the first two examples.  I'm more familiar with Darwin-SW, as 
I was involved in its development [7].  There is a Semantic Web Journal 
article about Darwin-SW [8], so I won't go into detail about it here, 
except to say that its data model was developed following an extensive 
discussion on the tdwg-content email list [9] about how members of the 
community understood the Darwin Core classes.  The relationship between 
Darwin-SW model and the historical 1993 ACS Model can be viewed at 
[10].  There are a bit over a million triples of data "in the wild" 
modeled on Darwin-SW in accordance with the DwC RDF guide, accessible at 
a SPARQL end point. [11]  Some examples showing how to play around with 
SPARQL queries of these data are at [12].

*The overlapping scope of Occurrence and Specimen types* - There is a 
long history behind the meaning of "Occurrence".  There is an 
out-of-date-summary of some of the discussion around this topic in the 
Darwin-SW documentation [13].  I think that at the time when Darwin Core 
was originally adopted, an Occurrence was considered a sort of 
superclass of the Specimen and Observation classes.  However, after a 
lot of discussion, the meaning of dwc:Occurrence was clarified by 
changing it to its current definition: "An existence of an Organism 
(sensu http://rs.tdwg.org/dwc/terms/Organism) at a particular place at a 
particular time."  In this view, an Occurrence isn't a concrete thing 
like a Specimen - it's more like a database join between an Event 
instance (time and place) and an Organism, which allows for a 
one-to-many relationship between a Organism and Occurrences, and a 
one-to-many relationships between an Event and Occurrences.  It also 
allows for a single occurrence of an organism at a time and place to be 
documented by one-to-many forms of evidence, which could include 
PreservedSpecimens, HumanObservation data, or images of various sorts.  
In RDF terms, an Occurrence could be thought of as a node that is linked 
to Event, Organism, and evidence instances nodes.  You can see this 
represented graphically at [7], where "dsw:Token" refers to a generic 
class for evidence.  In any case, separating Occurrence (as a node 
linking Events to Organisms) from Specimen allows an Occurrence to be 
documented by one to many instances of any kind of evidence, or even 
multiple kinds of evidence.  For example, an Occurrence could be 
documented by a PhysicalSpecimen as well as several images.  Here is an 
example of an organism with two Occurrences:
The first occurrence on 2013-07-24 was documented by 42 camera trap 
images, and the second occurrence on 2013-07-25 was documented by 21 
camera trap images.  You can see how this was represented in RDF at 
[14].  In most cases, specimen records will be much simpler than this, 
with one organism, documented at one occurrence, with evidence of one 
PreservedSpecimen.  Such a simpler case could be represented with a 
simpler model.  But the more complex model allows specimen-derived 
occurrence records to be merged with other kinds of occurrence records, 
such as the camera trap example I gave, mark-recapture bird banding 
observations, iNaturalist occurrences documented by photos of the 
organism, etc.

*Conflicting usage of Taxon fields in the Identification object* - In 
order to explain the rationale behind why what seem to be taxon-related 
properties are assigned to Identification instances, I must refer to the 
idea of "convenience terms" as expressed in Section 2.7 of the RDF 
Guide.[15]   In a perfect world, we would have the following:

a collection item linked by dwciri:inCollection to an IRI-identified 
an identification instance linked by dwciri:toTaxon to an IRI-identified 
taxon (a.k.a. taxon concept)
a location instance linked by dwciri:inDescribedPlace to an 
IRI-identified geographic place (a.k.a. "feature")

If the linked IRI-identified object resources were described by RDF, it 
would not be necessary to include any of the Darwin Core "convenience" 
properties included in Table 3.5 [16].  The information contained in the 
values of those properties could be discovered by dereferencing the 
object IRIs and traversing subsequent links from that RDF.  However, if 
those IRIs don't exist, then the convenience properties provide a 
string-based mechanism to relate the subject resource to other resources 
that should be linked to the same (unidentified) object resource.  So 
for example, if we say a specimen has the convenience properties and values

dwc:institutionCode ="MVZ"

we are not saying that "Mamm" is the collection code of the specimen and 
that "MVZ" is the institution code of the specimen.  Rather, we mean 
that the specimen should be linked to a collection (with unknown IRI) 
whose code is MVZ and whose owning institution has the code "MVZ".  
Similarly, if we say that an identification has the convenience 
properties and values


we are not saying that "yaeyamaensis" is the specific epithet of the 
identification and that "Hersiliiadae" is the genus of the 
identification.  Rather, we mean that the identification should be 
linked to a taxon (with unknown IRI) for which the specificEpithet part 
of its name string is "yaeyamaensis", which is included in the genus 
"Hersiliiadae".  This may seem odd, particularly if you are used to 
thinking of genus and specific epithet as properties of a taxon.  But 
the sets of DwC convenience properties are intended to be a temporary, 
string-based way to describe an unidentified resource to which the 
subject resource should be linked.  At some future time, if IRIs can be 
discovered, those sets of convenience properties might be dropped if 
dereferencing the IRIs provides the same information.  In these 
examples, one might replace with:

a collection item linked by dwciri:inCollection to 
an identification instance linked to 

although I don't think either of these IRIs currently dereference to 
meaningful machine-readable RDF (although they have human-readable web 

I hope that this has provided you with some answers, or at least a 
starting point for additional exploration or questions.  Please feel 
free to reply if there were parts of what I wrote that weren't clear.

Steve Baskauf

[1] https://github.com/tdwg/rdf
[2] http://groups.google.com/group/tdwg-rdf
[4] https://github.com/tdwg/vocab/blob/master/documentation-specification.md
[5] https://github.com/tdwg/vocab/blob/master/maintenance-specification.md
[6] https://github.com/tdwg/rdf/blob/master/DwCAncillary.md
[7] https://github.com/darwin-sw/dsw
[9] https://github.com/darwin-sw/dsw/wiki/TdwgContentEmailSummary
[10] https://github.com/darwin-sw/dsw/blob/master/img/acs-dsw-poster.pptx
[11] http://rdf.library.vanderbilt.edu/sparql?view
[13] https://github.com/darwin-sw/dsw/wiki/ClassOccurrence
[14] http://bioimages.vanderbilt.edu/org-jorgem/rec13_0004.rdf

Douglas Campbell wrote:
> Hi all,
> I am implementing Darwin Core in RDF as part of our API at Te Papa 
> (Museum of New Zealand).  My aim is to map our specimen metadata to 
> rich Darwin Core RDF using JSON-LD, then 'dumb down' to Simple Darwin 
> Core to contribute to virtual herbariums.  I have mocked-up some 
> records, however there are some areas where I'm not quite sure how to 
> interpret the Darwin Core RDF Guide.
> The areas of confusion I have include:
> * Best practice for connecting containers together
> * Dereferencing of the DwC IRI namespace
> * The overlapping scope of the Occurrence and Specimen types
> * Conflicting usage of Taxon fields in the Identification object.
> I'm hoping for suggestions:
> 1. Are there any implementations of DwC RDF data online that I could 
> look at as examples to follow?
> 2. What/to whom is the best way to ask specific questions about DwC RDF?
> At this stage our API prototype is only available internally but there 
> is some documentation available publicly at:
> https://github.com/te-papa/collections-api/wiki
> Thanks in advance,
> Douglas
> *Douglas Campbell*
> Business Analyst
> Collections Information Services
> Museum of New Zealand Te Papa Tongarewa
> ------------------------------------------------------------------------
> Visit the Te Papa website http://www.tepapa.govt.nz
> The email message together with the accompanying attachments may be 
> *CONFIDENTIAL*. If you have received this message in error, please 
> notify https://www.tepapa.govt.nz/about/contact-us/general-enquiries 
> immediately and delete the original message. The views expressed in 
> this message are those of the individual sender, except where the 
> sender specifically states them to be views of Te Papa.  Te Papa 
> employs strict virus checking measures and accepts no liability for 
> any loss caused either directly or indirectly by a virus arising from 
> the use of this message or any attached file.
> ------------------------------------------------------------------------
> This email has been filtered by SMX. For more information visit 
> smxemail.com <http://smxemail.com/>

Steven J. Baskauf, Ph.D., Senior Lecturer
Vanderbilt University Dept. of Biological Sciences

postal mail address:
PMB 351634
Nashville, TN  37235-1634,  U.S.A.

delivery address:
2125 Stevenson Center
1161 21st Ave., S.
Nashville, TN 37235

office: 2128 Stevenson Center
phone: (615) 343-4582,  fax: (615) 322-4942
If you fax, please phone or email so that I will know to look for it.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.tdwg.org/pipermail/tdwg-content/attachments/20160816/13333ca4/attachment.html>

More information about the tdwg-content mailing list