Print

Print


Yes, if you look at the W3C validation workshop I cited, much of the 
validation is done with SPARQL, with or without OWL. Using the closed 
world method may work in some cases, but it depends greatly on the 
nature of your closed and open worlds. What the Dublin Core AP work 
promotes is being very careful that your closed world validation needs 
do not affect your open world semantics. In other words, if you define 
your ontology with OWL constraints that you wish to interpret in a 
closed world, beware that those same constraints, with a different, open 
world context, apply to your data universally. This may not yield the 
desired results in the open world. (Note that many of the examples in 
the W3C workshop were enterprise uses of RDF, thus not terribly 
concerned with open world semantics. I think we have a special case with 
bibliographic data because not only do we want it to be open, but our 
data is, in essence, massively crowd-sourced, and not at all under a 
single point of control.)

In addition, there are many aspects of application profiles that are not 
part of the RDF concept. APs can define what LOV [1] calls a 
"vocabulary" -- that is, a set of classes and properties that are used 
in a particular data set.

In any case, if anyone wishes to participate, there is a new DCMI task 
group that over the next year will look into the needs for application 
profiles for metadata in RDF.[2]  It is just beginning, and anyone can 
join in. Discussion will take place on the dc-architecture [3] mailing 
list. If you have use cases to contribute, please send them to the list.

kc

[1] http://lov.okfn.org/dataset/lov/
[2] http://wiki.dublincore.org/index.php/RDF-Application-Profiles
[3] https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=dc-architecture&A=1

On 6/8/14, 9:31 PM, [log in to unmask] wrote:
> Perhaps we do not "constrain" in the usual data processing sense, but 
> useful information can be mechanically derived using these technologies.
>
> If I apply some ontology to a knowledge base and do not thereafter 
> discover a certain resource X categorized into a certain type T, I 
> normally interpret that in an open world as meaning that I just don't 
> yet know whether X is a T. But, for my own convenience and within the 
> privacy of my own systems, I _can choose_ to interpret that absence to 
> myself to mean that X is not a T, as long as I don't publish this 
> assertion or any assertions derived therefrom. It doesn't seem 
> reasonable to me to ignore this possibility for workflow.
>
> It is also very possible to reinterpret OWL with formal closed-world 
> semantics: Clark & Parsia 
> <http://docs.stardog.com/icv/icv-specification.html> offer a nice 
> example at: http://docs.stardog.com/icv/icv-specification.html
>
> ---
> A. Soroka
> The University of Virginia Library
>
> On Jun 7, 2014, at 4:58 PM, Karen Coyle <[log in to unmask] 
> <mailto:[log in to unmask]>> wrote:
>
>> There is, however, a bit of a problem with writing profiles and their 
>> constraints in RDF -- which is that RDF (and OWL) support inferences, 
>> but not constraints in the usual data processing sense. If you've 
>> looked into the W3C validation work, or the DCMI AP work, you see 
>> that there are points where one either bends the RDF rules, or moves 
>> into another language with different semantics.
>>
>> W3C validation: http://www.w3.org/2012/12/rdf-val/
>> DCMI RDF APs: 
>> http://wiki.dublincore.org/index.php/RDF-Application-Profiles
>> Shape Expressions: http://www.w3.org/2013/ShEx/Primer
>>
>> The DCMI group is going to be looking into APs from the DCMI point of 
>> view. I'm personally not sure what this is going to require, but in 
>> trying to think through how it might be defined in RDF, I run into 
>> problems, especially with cardinality and anything requiring structure.
>>
>> kc
>>
>> On 6/7/14, 7:14 PM, [log in to unmask] 
>> <mailto:[log in to unmask]> wrote:
>>> I have just learned about the Bibframe Profiles.
>>>
>>> Since Bibframe is based on RDF, why are Bibframe Profiles not based 
>>> on RDF?
>>>
>>> As I understand, the constraints described in Bibframe Profiles are 
>>> constraints on RDF elements and vocabularies, and do not constrain 
>>> mere data. I'm not sure about what is meant with "structural 
>>> constraints", maybe also integrity constraints?
>>>
>>> So expressing Bibframe Profiles in RDF as rules would be more 
>>> beneficial to the semantic web community.  I think it is possible to 
>>> express the rules as an ontology. By doing this, informal notations 
>>> or plain JSON or EBNF notations would no longer be necessary to 
>>> express a Bibframe Profile, the document could be rewritten to use 
>>> RDF (serialized in Turtle, JSON-LD, etc. whatever is convenient)
>>>
>>> Best,
>>>
>>> Jörg
>>
>> -- 
>> Karen Coyle
>> [log in to unmask] <mailto:[log in to unmask]> http://kcoyle.net
>> m: 1-510-435-8234
>> skype: kcoylenet
>

-- 
Karen Coyle
[log in to unmask] http://kcoyle.net
m: 1-510-435-8234
skype: kcoylenet