Re: Component attachment to data set level in OWL 2 DL

On 07/06/13 11:31, Markus Stocker wrote:
> On Fri, Jun 7, 2013 at 12:52 PM, Dave Reynolds
> <dave.e.reynolds@gmail.com> wrote:
>> On 07/06/13 10:25, Markus Stocker wrote:
>>>
>>> Hi Dave,
>>>
>>> Thanks for comments.
>>>
>>> On Fri, Jun 7, 2013 at 11:00 AM, Dave Reynolds
>>> <dave.e.reynolds@gmail.com> wrote:
>>>>
>>>> Hi Markus,
>>>>
>>>>
>>>> On 05/06/13 18:17, Markus Stocker wrote:
>>>>>
>>>>>
>>>>> Hi,
>>>>>
>>>>> The Example 4 [1] demonstrates how to attach a component at the data
>>>>> set level. However, qb:DataSet is a class and, as far as I understand,
>>>>> this statement is thus beyond OWL DL. Am I correct that this can be
>>>>> circumvented by using the Punning feature of OWL 2 DL in adding the
>>>>> assertion owl:NamedIndividual(qb:DataSet)?
>>>>
>>>>
>>>>
>>>> The Data Cube ontology is not OWL 2 DL in several ways and attachment
>>>> declarations is certainly one of them.
>>>
>>>
>>> Could it be of interest to document in what ways the ontology is not OWL 2
>>> DL?
>>
>>
>> Not sure.
>>
>> Most of the issues are minor details about the way the ontology is drafted,
>> rather than anything fundamental to the design. The component attachment
>> issue is the only substantive one and we've covered that.
>>
>> So in principle it would be only quite modest work to document and/or work
>> around the DL issues. However, I'm not sure there is any appetite for it. In
>> the couple of years it has been in use no one has reported any problems due
>> to the lack of DL compatibility. It is not the sort of vocabulary or
>> application area were inference beyond RDFS has much value. Given the
>> current slightly indeterminate status of the working group now is definitely
>> not the time to be adding work items, even small ones :)
>
> Yes, sounds reasonable. Thanks. I ran into this because I wanted to
> try the ontology and went ahead editing in Protege but, I believe, one
> needs these changes in order to do so. I will take a closer look at
> this.

OK, hadn't realized Protege was so picky these days.

The things to watch for are:
   * declares properties to be rdf:Property as well as e.g. 
owl:ObjectProperty, similarly for classes and rdfs:Class
   * no explicit imports so validators will complain about lack of 
declaration of things like skos:Concept
   * the qb:componentAttachment issue we've already mentioned

If you are working off a saved copy then worth refreshing, I fixed some 
typos this morning as a result of looking at it in response to your 
question.

Cheers,
Dave

Received on Friday, 7 June 2013 10:44:18 UTC