On 12/15/14 7:44 AM, Antoine Isaac wrote:
> Hi everyone,
>
>
>> The only "sort of" inferencing that we have in our use cases so far is
>> in R-64
>
> As Valentine noted, we suggest to consider the new R-224 as a
> 'replacement' for R-64 - at least one that matches better the need of
> the case and what Karen expressed.
> http://lelystad.informatik.uni-mannheim.de/rdf-validation/?q=node/451
>
>
> There is a couple of other requirements related to the (non) use of
> inference in the requirement on domain and ranges:
> http://lelystad.informatik.uni-mannheim.de/rdf-validation/?q=R-25-OBJECT-PROPERTY-DOMAIN
>
> http://lelystad.informatik.uni-mannheim.de/rdf-validation/?q=R-28-OBJECT-PROPERTY-RANGE
>
> See "Data not OK - example 2" in both, where the type of resources is
> missing in the original data.
> If one activates inference, then the error is not an error anymore
> because the type of the resource is likely to be added (I mean, if the
> property has been rightly defined in its ontology)
With regard to typing, W3C is assuming (AFAIK) that all types will be
explicit in the instance data - for the purposes of validation. There is
nothing, however, to prevent an application from running reasoning prior
to validation to add inferred triples to the data being validated. But
the general feeling right now is that validation acts on instance data
without requiring that the validator apply reasoning in order to do so.
As a practical matter, this makes sense to me. Can we adopt this for our
own work?
kc
>
> Cheers,
>
> Antoine
>
> On 12/13/14 5:52 PM, Karen Coyle wrote:
>> Which is all fine, except that to date we do NOT have such a use case
>> in our round-up of cases. So although there may be use cases that
>> SOMEONE has, we are not that SOMEONE.
>>
>> Now, this could be because most of the folks in our community haven't
>> had the opportunity to use inferencing, and therefore we haven't built
>> it into our data yet. However, since we are trying to K.I.S. (keep it
>> simple), I don't think that adding inferencing as a hypothetical will
>> help us meet our goals.
>>
>> The only "sort of" inferencing that we have in our use cases so far is
>> in R-64
>>
>> http://lelystad.informatik.uni-mannheim.de/rdf-validation/?q=node/71
>>
>> which says that if you have two properties, one which is a subproperty
>> of the other, and they have the same value, then you have redundancy
>> and one must be removed.
>>
>> That's it for using inferencing, as far as I can see. And I am seeing
>> that one as an edge case.
>>
>> kc
>>
>> On 12/13/14 8:24 AM, Bosch, Thomas wrote:
>>> I had the same idea a couple of month ago.
>>>
>>> I would like to state when inferencing should be done before actually
>>> validating the data and when inferencing is not wished for specific
>>> individuals.
>>> Therefore, I assign the class 'owl2spin:ToInfer' only to individuals
>>> for which inferencing should be performed:
>>>
>>> :hasDog rdfs:domain :Person .
>>> :Peter
>>> a owl:Thing, owl2spin:ToInfer ;
>>> :hasDog :Brian .
>>> :Brian
>>> a owl:Thing, owl2spin:ToInfer .
>>>
>>> The entailment would be:
>>> :Peter a :Person .
>>>
>>> Then the constraints associated with the class :Person are validated
>>> for :Peter as well.
>>>
>>> The power of this is that you do not have to assign individuals to
>>> specific classes explicitly in your data. Class assignments are
>>> inferred and then constraints associated with these classes are
>>> validated as well.
>>>
>>>
>>> Cheers,
>>> Thomas
>>>
>>>
>>> Here's a use case for inference:
>>>
>>> A person has two distinct biological parents.
>>> schema:
>>> sh:Person { :hasBiologicalParent @sh:Parent{2} }
>>> sh:Parent { foaf:mbox IRI }
>>> violating data:
>>> <Alice> :hasBiologicalParent [ foaf:mbox <mailto:[log in to unmask]> ],
>>> [ foaf:mbox <mailto:[log in to unmask]> ].
>>> An OWL IFP asserts that there's only one biological parent above.
>>>
>>> and here's a use case where inference hides an error
>>>
>>> A person must be so labeled.
>>> schema:
>>> sh:Persin { a (foaf:Person) }
>>> violating data:
>>> <Bob> foaf:name "Bob".
>>> Domain inference will conclude that <Bob> is a person.
>>>
>>> At a clinical data hackathon last week, someone proposed adding
>>> control to the schema langauge itself, à la:
>>> sh:Person { :hasBiologicalParent INFER(@sh:Parent){2} }
>>> to permit inference on the object, or
>>> sh:Person { INFER(:hasBiologicalParent) @sh:Parent{2} }
>>> to permit (subproperty) inference on the predicate.
>>>
>>>
>>> thoughts?
>>>
>>>
>>> * Karen Coyle <[log in to unmask]> [2014-12-13 07:13-0800]
>>>> Right, but we're talking here about validation, not use of RDF in
>>>> the open world.
>>>>
>>>> kc
>>>>
>>>> On 12/13/14 1:26 AM, Bosch, Thomas wrote:
>>>>> As domain and range inferencing is part of the very very basic
>>>>> inferencing mechanism of RDFS this seemed to me as it should /
>>>>> could be interesting for the DCMI community.
>>>>> I'm not that deep in the use cases.
>>>>>
>>>>>
>>>>> Kind regards,
>>>>> Thomas
>>>>>
>>>>>
>>>>> Thomas, so far I haven't seen anything in our use cases that refers to
>>>>> domain and range inferencing. Let me know if I missed that.
>>>>>
>>>>> kc
>>>>>
>>>>> On 12/12/14 9:49 AM, Bosch, Thomas wrote:
>>>>>> You may also consider property domain and range inferencing.
>>>>>>
>>>>>>
>>>>>> Thomas
>>>>>>
>>>>>>
>>>>>> W3C is starting to fill in pages with requirements. In this page,
>>>>>> I try
>>>>>> to explain the DC approach to inferencing. Please let me know if you
>>>>>> think I got it wrong.
>>>>>>
>>>>>> https://www.w3.org/2014/data-shapes/wiki/ISSUE-1:_What_inferencing_can_or_must_be_used#DC_Application_Profiles
>>>>>>
>>>>>>
>>>>>> Thinking it over, it does occur to me that for the few cases that we
>>>>>> have that look at sub/super we could create a requirement that
>>>>>> does not
>>>>>> have to read the ontology, such that
>>>>>>
>>>>>> if [specific title] and [dc:title] use [specific title]
>>>>>>
>>>>>> (I may have the backward from the actual case.)
>>>>>>
>>>>>> It seems to me that, especially for profiles using a mix of
>>>>>> ontologies,
>>>>>> having to have access to all of the ontologies in order to do
>>>>>> validation
>>>>>> could be a huge burden.
>>>>>>
>>>>>> kc
>>>>>> --
>>>>>> Karen Coyle
>>>>>> [log in to unmask] http://kcoyle.net
>>>>>> m: 1-510-435-8234
>>>>>> skype: kcoylenet/+1-510-984-3600
>>>>>>
>>>>>
>>>>> --
>>>>> Karen Coyle
>>>>> [log in to unmask] http://kcoyle.net
>>>>> m: 1-510-435-8234
>>>>> skype: kcoylenet/+1-510-984-3600
>>>>>
>>>>
>>>> --
>>>> Karen Coyle
>>>> [log in to unmask] http://kcoyle.net
>>>> m: 1-510-435-8234
>>>> skype: kcoylenet/+1-510-984-3600
>>>
>>> --
>>> -ericP
>>>
>>> office: +1.617.599.3509
>>> mobile: +33.6.80.80.35.59
>>>
>>> ([log in to unmask])
>>> Feel free to forward this message to any list for any purpose other than
>>> email address distribution.
>>>
>>> There are subtle nuances encoded in font variation and clever layout
>>> which can only be seen by printing this message on high-clay paper.
>>>
>>
>
--
Karen Coyle
[log in to unmask] http://kcoyle.net
m: 1-510-435-8234
skype: kcoylenet/+1-510-984-3600
|