Re: looking for a social schema

Le 18/11/2011 16:48, Paul Gearon a écrit :
> Hi Heiko,
>
> On Fri, Nov 18, 2011 at 4:33 AM, Heiko Paulheim
> <paulheim@ke.tu-darmstadt.de>  wrote:
>> Due to the open world assumption, the value of that counter would (probably)
>> we wrong.
>>
>> If I have
>> :Peter foaf:knows :Stephen .
>> :Peter foaf:knows :Marc .
>>
>> and, based on that knowledge, I added
>>
>> :Peter myschema:friendnb "2"^^xsd:integer .
>>
>> this would not be a valid conclusion - there may be a lot more friends of
>> Peter which are not in my knowledge base (don't we all have friends without
>> a facebook account?), and :Stephen and :Marc might even refer to the same
>> person. In other words, with that approach, I would add knowledge to my
>> knowledge base which is potentially wrong.
> While you are correct in saying that the inference of 2 friends is
> invalid, the idea of cardinality is not inconsistent with the open
> world assumption (OWA). Melvin's original question was about a
> property that can be used to declare cardinality, and this is fine
> with the OWA. Indeed, OWL uses it.
>
> So, for instance, if you declare:
>
>   :Peter myschema:friendnb "2"^^xsd:integer .
>
> and then you say:
>
>   :Peter foaf:knows :Stephen .
>   :Peter foaf:knows :Marc .
>   :Stephen owl:differentFrom :Marc .
>
> Then you know that we have identified all the friends of :Peter. This
> does not preclude another statement of the form:
>
>   :Peter foaf:knows :Steve .
>
> But since we already knew all of :Peter's friends, then we know that
> this new statement must refer to an alias for one of the existing
> friends.
>
> Taking it further:
>
>   :Steve owl:differentFrom :Marc .
>
> Means that:
>
>   :Steve owl:sameAs :Stephen .
>
> All of this is just a long-winded way of explaining that a cardinality
> predicate is not in conflict with the open world assumption. However,
> it works with different use cases than with the closed world.
> Specifically, under the OWA you cannot derive the current cardinality,
> but you can declare it.
>
> Regards,
> Paul Gearon
>

So i could have an "Owner" (of an account) class and a "Friends" class 
which are subclasses of an union of both
and an inverse hasFriends - isFriendOf and a value restrictions right?

Cheers
Ghalem

Received on Friday, 18 November 2011 17:21:02 UTC