I don't want to put the cart before the horse here because the Darwin Core RDF Guide has not been formally introduced as an addition to Darwin Core (it's waiting until John W. feels the time is right to do so in the context of dealing with the various issues he's been working through).  But it is in the queue with recommendation of the RDF Task Group to adopt and can be viewed online.  I mention this because it contains a specific recommendation for how to deal with terms that have multiple values in a concatenated list.  See
http://code.google.com/p/tdwg-rdf/wiki/DwcRdfGuideProposal#2.5.1_Definition_of_dwcuri:_terms
for the details. 

In a nutshell, the guide establishes the convention that the existing dwc: namespace terms be used with literals which are formatted as described in the existing standard (e.g. a delineated, concatenated list).  It creates new versions of the terms (in a new namespace dwcuri:) which are intended to be repeatable and to have single values which are URI references. 

I am hesitant to bring this up because the guide has not been formally introduced nor has a 30 day discussion period been declared.  But I think in light of this discussion, it is important for people to know that the guide does address this issue in the context of RDF. 

Steve

Markus Döring wrote:
+1 to remove multi value recommendations from the main DwC definitions and leave it to the implementations to deal with lists if needed.

As many terms currently are in plural we can easily create new terms for single values. I am not entirely convinced though that these terms are very useful if they combine various properties into an unstructured string. In these cases it might be better to define several single value terms instead. A quick attempt to create single value terms:


Terms currently in plural form:
----------------------------------------
dataGeneralizations
-> dataGeneralization

dynamicProperties
-> dynamicProperty

preparations
-> preparation

associatedSequences
-> associatedSequence

georeferenceSources
-> georeferenceSource

associatedReferences
-> associatedReference  (id and/or citation string ?)

otherCatalogNumbers
-> otherCatalogNumber   (needed at all if there is already catalogNumber ?)

previousIdentifications
-> previousIdentification  (combines various identification properties into human string. Redefine as just the previously identified scientificName or deprecate entirely ?)

associatedOccurrences
-> associatedOccurrence  (combines occurrenceID with relation type. Maybe just associatedOccurrenceID ? Or just deprecate it in favor of the ResourceRelationship terms ?)

associatedTaxa
-> associatedTaxon  (combines taxonID or scientificName with associationType)



Terms already in singular form. 
------------------------------------
Can we redefine those terms or do we need to create new ones?

typeStatus
# seems to combine any property about typification/type designation right now, although I have mostly seen a single status so far.
The discussion page recommends values for the "status portion of the content". How about restricting the term to this status / kind of type?

vernacularName
# I have not seen anyone using this as a list. Is anyone aware of such a case? Might not cause too much trouble to redefine

recordedBy

associatedMedia   

higherClassification   
# isn't a single classification always a list? 

higherGeography

informationWithheld



Markus




On 07.10.2013, at 18:15, joel sachs wrote:

  
On Mon, 7 Oct 2013, John Wieczorek wrote:

    
On Mon, Oct 7, 2013 at 4:54 PM, Tim Robertson [GBIF]
<trobertson@gbif.org> wrote:
      
I kind of expected it was futile to make the plea "Please ignore the
issue of whether the idea of list-type terms is a
good idea or not - that is not the issue we're trying to resolve
here." I had to try.
          
Sorry John, I fell into your trap.
        
:-) You were not alone.

      
Surely though, talking about solutions to problems rather than the problem itself is the _worst_ option available to us?
        
I just wanted to keep the issues separate and focus on the one
submitted, for the very reason that it will otherwise get to broad and
contentious to provide any solutions at all. I don't much like
spending energy when that is the likely outcome. The process more
often seem to yield results when it is kept simple. And in this case,
having a better recommendation does not set us in any worse position
than we are now. No one presented an issue to the tracker recommending
the deprecation of all "list" terms".
      
I do plan on submitting an issue to the tracker. It won't be to deprecate the terms, but, as Tim suggests to change the definitions so that the recommendations on how to deal with multiple values is left to the various representation guides (text, xml, rdf). (If anyone else wants to submit this issue first, please go ahead.)

I'm glad that Tim fell into your trap, as it raises an important issue, withour precluding Darwin Core (through the representation guides) from providing consistent guidance on these terms.

Joel.







    
Consider the likes of this list term
http://rs.tdwg.org/dwc/terms/#typeStatus
The description suggests a separated and concatenated list but the example
(unless I misunderstand) is showing only 1 list item which is a triplet of
"type + author + pub" in a human readable form.  This one field is actually
suggesting a structure of a repeatable triplet, so need 2 delimiters if
machines are to extract the scientific name for the typification .   Perhaps
these terms are really just verbatim text blocks intended for human
consumption (which is fine with me, and we don't need to define delimiters)?
Or perhaps we should be discussing terms to atomize them further (e.g.
introduce dwc:typeName and dwc:typePublication)?
        
Yes, the example gives only one typeStatus entry, not a list. Yes, one
can argue that the content mixes concepts if those distinct concepts
are of interest. A look at the history of typeStatus will reveal that
it has its origins deep in the Darwin Core history, and no one has yet
suggested that it should be other than what it is. Another item for
the issue tracker if anyone wants to defend a change.

      
If we are heading this way though, can I also suggest we consider declaring
the expected ordering on lists where omitted?  The likes of
http://rs.tdwg.org/dwc/terms/#higherGeography doesn't  have one whereas
http://rs.tdwg.org/dwc/terms/#higherGeography does.
        
Same example. Did you mean
http://rs.tdwg.org/dwc/terms/#higherClassification? It would be a fine
thing to amend the recommendation for higherGeography to suggests the
ordering. If anyone seconds the motion I'll create an issue for it.

      
There are definitely more rigorous ways to share the information than
in concatenated lists. The "list" terms are just as Tim describes, an
attempt to share in a flat data structure data that do not fit well in
a flat structure, but are nevertheless of common interest. There
probably shouldn't be an expectation that one could process the
content of such fields and derive individual values, as we can't even
get simple content under control yet (see
http://soyouthinkyoucandigitize.wordpress.com/2013/07/18/data-diversity-of-the-week-sex/).

Nevertheless, these terms do exist, and they expect lists, and people
are using them in distinct ways that make them a challenge to process.
It would be nice to give guidance. I have no problem if that guidance
stays out of the term definitions, but we have a legacy problem of
definitions that tell us that the content should consist of a
delimited list.
          
Other than deprecating and redefining as new concepts (terms), I don't see
any robust way I am afraid.  Some things are just not meant to be
denormalized.
        
That would be a fine conclusion as well, if we can get consensus. I
would then just add secondary documentation saying "Beware all ye who
enter (data) here."

      
Cheers,
Tim





On Mon, Oct 7, 2013 at 4:02 PM, Tim Robertson [GBIF]
<trobertson@gbif.org> wrote:

I suspect any attempt to find a universal delimiter will be flakey at best -

see Unicode character 1 as an example [1].

I would urge DwC to stop at only defining the concept of each term and leave

it to the serialization formats, schema definitions, data models etc (e.g.

DwC-A, XML, RDF, JSON, HTML, excel templates etc) to define those kind of

things.


If you were to design an XML schema you would use things like:


<tim:identifications>

<dwc:scientificName>A</dwc:scientificName>

<dwc:scientificName>B</dwc:scientificName>

<dwc:scientificName>C</dwc:scientificName>

</tim:identifications>


and not:


<tim:identifications>

<dwc:scientificName>A|B|C</dwc:scientificName>

</tim:identifications>


I don't think it wise for the DwC standard to suggest anyone should.


I suspect this request stems from those working with denormalized data

structures, and trying to shoe-horn all data into flat structures (e.g.

DwC-A).  I think that is a dangerous path to go down, and makes things more

difficult for both producers and consumers.  Very quickly you will get into

the situation where you will want to also suggest "well the element at index

[0] of field X should be interpreted as the index [0] for field Y" (e.g.

identifications and identification dates).


Cheers,

Tim


[1] http://www.fileformat.info/info/unicode/char/1f/index.htm





On Oct 7, 2013, at 3:45 PM, Steve Baskauf wrote:


I don't have an opinion about what the recommended delimiter should be, but

I think it would be beneficial for there to be consistency between Darwin

Core and Audubon Core.  You can see what the recommendation is for Audubon

Core at

http://terms.gbif.org/wiki/Audubon_Core_%281.0_normative%29#Lists_of_plain_text_values

- it's the pipe "|".  Either Darwin Core should go with this, or if there is

a consensus reached here that is different, then AC should be changed before

it is ratified, which potentially could happen in a matter of weeks.  It is

highly likely that there will be records that are a mixture of AC and DwC,

so it would not be a good thing for the recommendations to differ.


Steve


Markus Döring wrote:


Hi John et al.,


I would like to see a single recommended default delimiter, preferrably the

semicolon as its natural and hardly used in values.

For dwc archives there is a multiValueDelimiter attribute for every term

mapping that allows to declare other delimiters if needed.


Currently it is hardly possible to detect multi values in a field and you

can just test for some often used ones but even then you never know if they

were meant to be delimiters.

Having a single default value helps to get the idea of multi values across

and make it a bit more accessible I believe.


dwc:vernacularName I would personally prefer to see as a single value term

as it is mostly useful in combination with a locale and rarely is shared on

its own.

Seeing dwc:typeStatus being a multi value term also feels wrong as the name

is in singluar while the others carry the multi value nature in the name

already.



Markus




n 07.10.2013, at 12:28, John Wieczorek wrote:




Dear all,


On the list of pending Darwin Core issues is a topic of general

concern about terms that could or do recommend the concatenation and

delimiting of a list of values. The specific issue was submitted on

the Darwin Core Project site at

https://code.google.com/p/darwincore/issues/detail?id=168. Right now

there is variation in the recommendations of distinct terms.


The Darwin Core terms that could be used to hold lists include the

following (use the index at

http://rs.tdwg.org/dwc/terms/index.htm#theterms to find and see the

details of each of these):


informationWithheld

dataGeneralizations

dynamicProperties

recordedBy

preparations

otherCatalogNumbers

previousIdentifications

associatedMedia

associatedReferences

associatedOccurrences

associatedSequences

associatedTaxa

higherGeography

georeferenceSources

typeStatus

higherClassification

vernacularName


There are some issues. Many terms do not show examples. Most of those

that do show examples recommend semi-colon (';') -

associatedOccurrences, recordedBy, preparations, otherCatalogNumbers,

previousIdentifications, higherGeography, georeferenceSources, and

higherClassification, The example for higherClassification does not

have spaces after the semi-colon while all others do.


Terms that could hold a list of URLs would require a delimiter that

would be an invalid part of a URL unless it was escaped. This

precludes comma (','), semi-colon (';'), and colon (':'), among

others. One possibility here might be the vertical bar or "pipe"

('|').


The term dynamicProperties is meant to take key-value pairs. The

examples suggest the format key=value, with any list delimited by a

semi-colon, for example, "tragusLengthInMeters=0.014;

weightInGrams=120". The example for associatedTaxa also shows a

key-value pair ("host: Quercus alba"), but it is formatted differently

from the examples for dynamicProperties. There are other terms, such

as vernacularName, which could potentially also take a key-value pair,

though it is not currently recommended to be a list.


Please ignore the issue of whether the idea of list-type terms is a

good idea or not - that is not the issue we're trying to resolve here.

Instead, the issue is whether a consistent recommendation can be made

for how to delimit the values in a list. And if not a consistent

recommendation, can we make specific recommendations for distinct

terms? If specific recommendations can be made for a term, should that

be reflected in examples within the term definitions, or should such

recommendations reside only in Type 3 supplementary documentation such

as that which can be found on the Darwin Core Project site at, for

example,

https://code.google.com/p/darwincore/wiki/Occurrence#associatedSequences?

Should some of these terms have specific recommendations to contain

only single values (e.g., vernacularName), in which case they are not

really viable in Simple Darwin Core?


Cheers,


John

_______________________________________________

tdwg-content mailing list

tdwg-content@lists.tdwg.org

http://lists.tdwg.org/mailman/listinfo/tdwg-content



_______________________________________________

tdwg-content mailing list

tdwg-content@lists.tdwg.org

http://lists.tdwg.org/mailman/listinfo/tdwg-content


.





--

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.

http://bioimages.vanderbilt.edu


_______________________________________________

tdwg-content mailing list

tdwg-content@lists.tdwg.org

http://lists.tdwg.org/mailman/listinfo/tdwg-content




_______________________________________________

tdwg-content mailing list

tdwg-content@lists.tdwg.org

http://lists.tdwg.org/mailman/listinfo/tdwg-content




        
_______________________________________________
tdwg-content mailing list
tdwg-content@lists.tdwg.org
http://lists.tdwg.org/mailman/listinfo/tdwg-content
      
_______________________________________________
tdwg-content mailing list
tdwg-content@lists.tdwg.org
http://lists.tdwg.org/mailman/listinfo/tdwg-content
    

_______________________________________________
tdwg-content mailing list
tdwg-content@lists.tdwg.org
http://lists.tdwg.org/mailman/listinfo/tdwg-content

.

  

-- 
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.
http://bioimages.vanderbilt.edu