Re: A solution to integrate CWA into OWA.

Hi Dave, may I point out that your suggestion won't quite work, e.g.,  
in cases where we know that

John is an instance of ((R some C) or (P some D))....

In particular, your proposal only works for so-called Horn Logics,  
i.e., one where for each consistent ontology, we have a single  
'canonical' model against which we can query. In general, we have to  
consider numerous of those models....

As an alternative, I'd suggest to go with (2): extend DL-query with  
(epistemic) 'Not2' so that we can distinguish e.g., between

- people who are known not to have any children and
- people who are not known to have any children.

As the above a clearly two different concepts, it should be acceptable  
to have to different queries for them -- so as to not confuse them!

Cheers, Uli

On 2 Mar 2011, at 10:36, Dave Andersen wrote:

> Dear All,
>
> Hereby I propose a SOLUTION to elegantly integrate CWA in OWA and  
> put an END to a (I think for MANY people) very frustrating  
> situation: a situation where one can NOT find OWL indivuals that  
> LACK certain properties just with the simple boolean algebra term  
> "NOT" (due to OWA) .
>
> PROPOSAL:
> A CLOSURE AXIOM on the Knowledge Base {KB} that the user wants to  
> query (similar to exhaustive class {enumeration} or  
> owl:allValuesFrom).
>
> THIS IS HOW IT WORKS:
> User can activate OPTION to let reasoner put a CLOSURE AXIOM on KB:  
> {KB}.
> 1. The reasoner follows the normal procedure to infer statements  
> from asserted statements.
> 2. The reasoner now applies a CLOSURE action on the instantiated  
> individuals and their properties as follows:
>     2.1 Collect all the individuals and the properties (ONLY their  
> NAMES, values are NOT important)*.
>     2.2a Create as many so called "NullProp" classes as there are  
> property(names).
>     2.2b Give each NullProp class exactly 1 single statement:  
> property exactly 0. All properties now also have as many NullProp  
> classes.
>     2.3 For every individual check that for the property at least 1  
> statement is present.
>     2.4a If it does, then move to the next property and repeat  
> procedure.
>     2.4b If it doesn't, then make this individual a member of the  
> NullProp class for this property.
>     2.4c Move to the next property and repeat procedure.
>     2.5 Are all properties assigned to this individual? Then move to  
> the next individual and repeat procedure.
>     3. At the end ALL individuals MUST HAVE ALL properties assigned  
> to, either asserted, inferred or via CLOSURE action (i.e. property  
> exactly 0).
>
> RESULT:
> 1. Every query can now be carried out with simple boolean algebra to  
> find Individuals with OR WITHOUT certain properties: e.g. DL-Query  
> with simple Manchester Syntax (incl. "NOT").
> 2. NO more complicated SPARQL or (worse) PROPRIETARY queries!
> 3. Simple constraints checking!!
>
> Since (a part of) set theory is already teached in K12 schools, we  
> greatly  increases our potential (future) OWL userbase. E.g. in  
> social networks and messages you look for SPECIFIC things like  
> persons,interests, opportunities, files... OR the LACK thereof, and  
> how to carry this out in a SIMPLE way.
>
> User Axiom :-) : To make things WIDELY accepted, LOGICAL right is  
> NOT enough, it must also be PRACTICALLY right (read: relatively  
> SIMPLE in use).
>
> note *: an instantiated set of individuals and their properties can  
> be regarded as a class. So one actually applies a closure axiom on  
> the (new) instantiated class, right?
>
> How about this? :-)
>
> Regards,
> DJ