Re: ISSUE-65 (excess vocab): REPORTED: excessive duplication of vocabulary

Jim Hendler wrote:
> Le me be clear, I didn't agree I liked this solution, I agreed it was
> the minimal solution under current constraints :-)  
> What I do agree with Boris on is that I see no reason to allow punning
> of object and datatype properties - these are inherently different than

Well... I did meet one example. DCMI (the organization behind the Dublin
Core metadata) is having problems exactly on that. They have an abstract
model document[1] where they speak about 'value surrogate' that can
either be a literal or non-literal. When mapping this abstract model to
RDF[2] they hit this problem (eg, is the value of a dcterm:subject
property a literal or not).

_At the moment_ their terms are not yet defined in OWL, but they were
thinking about it. The strict separation of datatype and object
properties was a problem they already realized they have (they are not
really happy having two different property names for 'subject' for
example). Punning may be an answer to their concerns.

Ivan



[1] http://dublincore.org/documents/abstract-model/
[2] http://dublincore.org/documents/dc-rdf/



> the class/instance punning, which is based on a known KR feature
> (metamodeling - which got its own WG Note from the dissemination SWBPD
> WG).  Again, I've yet to see any use case where this kind of punning is
> useful - I have seen some in some advanced KR systems where morphing
> from one to the other is doable, but I don't see us taking OWL anywhere
> near that far into pi-logic space.
> 
> On Nov 21, 2007, at 4:31 AM, Boris Motik wrote:
> 
>>
>> Hello,
>>
>> I do agree that duplication of vocabulary is rather nasty. I've spoken
>> to Jim Hendler at ISWC about it, and we came to a possible
>> solution, which I'd like to overview next. Before giving the solution,
>> however, let me just recapitulate why we introduced the
>> duplicated vocabulary in the first place.
>>
>> 1. OWL 1.1 allows the same URI to be used as both an object and as a
>> data property. (OWL 1.1 in general allows for punning; however,
>> for this particular problem, only punning on object and data
>> properties is a problem. Thus, let us not discuss here the general
>> problems of punning in OWL 1.1 - this should be discussed separately.)
>>
>> 2. Parsing OWL RDF is difficult: one needs to first scan the file for
>> the appropriate rdf:type triples, after which one needs a
>> second pass to actually output the axioms. Assuming that you parse
>> just one ontology, this is a pain but not a serious problem;
>> however, ontology imports exacerbate this problem. Assume that you are
>> parsing an ontology O that imports an ontology O'.
>> Furthermore, let us assume that O contains, for example, a
>> someValuesFrom restriction on the property p. In OWL 1.0 it can happen
>> that the triples in O do not allow us to disambiguate the type of p;
>> thus, we need to look at the imported ontology O' to find out
>> what the correct type of p is. This makes parsing of OWL RDF really
>> difficult: you can't process an ontology by itself, but you need
>> to look at the imported ontologies as well.
>>
>> Even worse, what if the imported ontology O' is not in OWL RDF but in
>> some other format? (For example, KAON2 allows a file ontology
>> to import an ontology that resides in a relational database.) Parsing
>> is now next to impossible. Thus, to allow parsing an ontology
>> O by looking only at the triples in O, we introduced the typed vocabulary.
>>
>>
>> ---------------------------------------------------------------
>>
>> And now for the solution.
>>
>> I do agree that the first point is not really a use-case: I do not
>> expect that users will actually want to use the same URI as both
>> an object and a data property. In contrast, being able to parse each
>> ontology by itself seems like a desirable property that should
>> be preserved. Hence, I propose to change the specs as follows:
>>
>> 1. I would leave the structural specification as it is. In this spec,
>> I would allow the same URI to be used as both an object and a
>> data property, and I would leave the typed vocabulary as it is. This
>> is not so much driven by the desire for punning; rather, the
>> structural specification is intended to provide guidance for
>> implementors of OWL APIs. In all APIs I know of, you do have a
>> separation of object and data properties; therefore, we should keep
>> this separation in the spec as well. Whether you then allow the
>> same URI to be used as an object and as a data property is then really
>> irrelevant for all intents and purposes.
>>
>> 2. A transformation of an ontology from the structural format (i.e.,
>> from the functional-style syntax) into RDF should be possible
>> only if you strictly separate the properties into object and data ones.
>>
>> 3. The transformation into RDF would then be roughly the same as in
>> OWL 1.0, with the following difference: the resulting RDF graph
>> would be *required* to type all properties used in the graph.
>>
>> 4. An RDF graph G could be parsed into the structural format only if
>> each URI that is used as a property is correctly typed IN THIS
>> GRAPH. If, for example, some URI p is used in a someValuesFrom
>> restriction but G contains neither <p rdf:type owl:ObjectProperty>
>> nor <p rdf:type owl:DatatypeProperty>, then G would not constitute a
>> valid OWL 1.1 DL ontology.
>>
>>
>> This solution seems to have the benefit of satisfying everyone: we can
>> parse each RDF graph by itself, there is no typed vocabulary,
>> and most reasonable use cases seem to be satisfied. Let me know how
>> you feel about this solution.
>>
>> Regards,
>>
>> Boris
>>
>>
>>> -----Original Message-----
>>> From: public-owl-wg-request@w3.org
>>> [mailto:public-owl-wg-request@w3.org] On Behalf Of OWL Working
>>> Group Issue Tracker
>>> Sent: 20 November 2007 14:49
>>> To: public-owl-wg@w3.org <mailto:public-owl-wg@w3.org>
>>> Subject: ISSUE-65 (excess vocab): REPORTED: excessive duplication of
>>> vocabulary
>>>
>>>
>>>
>>> ISSUE-65 (excess vocab): REPORTED: excessive duplication of vocabulary
>>>
>>> http://www.w3.org/2007/OWL/tracker/issues/
>>>
>>> Raised by: Jeremy Carroll
>>> On product:
>>>
>>>
>>> The member submission documents seem to replace a good many
>>> properties from OWL 1.0 with three
>>> properties in OWL 1.1. (The old version, and two new versions, one
>>> for data properties, and one for
>>> object properties)
>>>
>>> This:
>>> - creates additional work for implementors
>>> - creates additional work for documentation writers
>>> - potentially creates confusion for people as they learn the language
>>>
>>> Do the benefits outweight these (and other) costs?
>>>
>>>
>>>
>>>
>>>
>>
>>
>>
> 
> "If we knew what we were doing, it wouldn't be called research, would
> it?." - Albert Einstein
> 
> Prof James Hendler http://www.cs.rpi.edu/~hendler
> Tetherless World Constellation Chair
> Computer Science Dept
> Rensselaer Polytechnic Institute, Troy NY 12180
> 
> 
> 
> 

-- 

Ivan Herman, W3C Semantic Web Activity Lead
Home: http://www.w3.org/People/Ivan/
PGP Key: http://www.ivan-herman.net/pgpkey.html
FOAF: http://www.ivan-herman.net/foaf.rdf