RE: OWL-1.1-Full TF [Was: Introductions]

Hi, Jeremy!

Jeremy Carroll wrote:

>I was intending to do this tomorrow morning - with both syntaxes (as 
>suggested by Peter and Pat), but probably limited to only min or max.

For Peter's variant of a minimum object-QCR, along the lines of table "Conditions on OWL restrictions" in sec. 5.2 of the S&AS [0]:

  IF
    <x,n> IN EXT_I(S_I(owl11:minObjectQualifiedCardinality)) AND
    <x,p> IN EXT_I(S_I(owl:onProperty)) AND
    <x,c> IN EXT_I(S_I(owl11:onClass))
  THEN
    x IN IOR,
    n IN LV, 
    n is a non-negative integer,
    c IN IOC,
    p IN IOOP,
    CEXT_I(x) = {u IN IOT | card({v IN IOT: <u,v> IN EXT_I(p) AND v IN CEXT_I(c)}) >= n }  

with

  IOR  : set of all restrictions
  LV   : literal values
  IOC  : classes
  IOOP : object properties (in Full just all properties)
  IOT  : individuals (in Full just all resources)

Note the differences compared to 'minCardinality' in the table:

  * p is only in IOOP, not in IOOPuIODP, since we have an /object/ QCR. Of course, for Full, this is out of relevance, since all data properties are object properties (note that the tables in sec. 5.2 are generic for Full *and* the DL characterization in sec. 5.4).

  * the bound variable v is only in IOT, not in IOTuLV, for the same reason as above

  * the set specification in the 'card(.)' expression contains an additional condition "AND v in CEXT(c)". That's the "Q" in "QCR".

For the semantical difficulties with Pat's approach, see my mails in public-owl-dev. Actually, Pat had /two/ approaches. The first one had inherent problems, which cannot be resolved [1]. The second one would need a change in the semantic conditions for cardinalities in comparision to 1.0-Full [2] (this was Pat's answer to my answer to his second approach: both mails before went accidentally offlist). The problem is that the semantic conditions for cardinality restrictions fire already on

  x owl:*Cardinality n
  x owl:onProperty p

without taking an additional

  x rdf:type owl:Restriction 

into account. Changing the semantic conditions in a way that all three triples are necessary to fire would result in a (probably not very important, as Pat claimed [2]) change in semantics, because for certain RDF graphs there would then be entailments missing, namely those graphs which do not use 'owl:Restriction' triples in cardinality restrictions.

>> since it is really something like an "implementation detail" 
>of the OWL-Full
>> spec (I do not find a better name at the moment). 
>
>Yes - but isn't producing an OWL Full semantics achieved by doing all 
>these implementation details.
>
>> Otherwise, we could bring
>> up a big list of actions like:
>> 
>>   "Attempt Wiki sketch of ?x semantics OWL-Full"  
>> 
>> where '?x' stands for
>> 
>>   * sub property chains
>>   * asymmetric properties
>>   * self restrictions
>>   * data ranges
>>   * ...
>> 
>
>...
>
>I think that is the task. And QCRs are as good a place to start as any 
>other. Getting one or two of these items done, might then give 
>us enough 
>idea of where we might be going to have an informed discussion.

Yes. The question is whether this should in any case become a WG action. It has its pro's, of course, since it makes our process transparent to the WG, and it defines clear targets with clear time limits. But some people might perhaps think that by this our process get's a bit *too* transparent. :) This question has to be discussed by the WG.

>I would think subproperty chains were fairly important, with 
>some of the 
>things as more minor.

I also have the feeling that subproperty chains are the most interesting feature for OWL-Full modelling, and I have started to work on a an analysis. I will discuss it here, when I have finished my analysis. 

>What is your sense of what the major issues are, that are not 
>reducible 
>to a collection of implementation details?

There might at least be one big and fundamental issue, which I have to check first. Please be patient. :)

>Bijan's alternative proposal is probably one ...
>
>I think whether we wish to revisit the layering issues is another (and 
>this say interacts with Bijan's skolemization proposal).

I will first need to have a closer read of this.

Cheers,
Michael

[0] <http://www.w3.org/TR/owl-semantics/rdfs.html#5.2>
[1] <http://lists.w3.org/Archives/Public/public-owl-dev/2007OctDec/0248.html>
[2] <http://lists.w3.org/Archives/Public/public-owl-dev/2007OctDec/0250.html>

--
Dipl.-Inform. Michael Schneider
FZI Forschungszentrum Informatik Karlsruhe
Abtl. Information Process Engineering (IPE)
Tel  : +49-721-9654-726
Fax  : +49-721-9654-727
Email: Michael.Schneider@fzi.de
Web  : http://www.fzi.de/ipe/eng/mitarbeiter.php?id=555

FZI Forschungszentrum Informatik an der Universität Karlsruhe
Haid-und-Neu-Str. 10-14, D-76131 Karlsruhe
Tel.: +49-721-9654-0, Fax: +49-721-9654-959
Stiftung des bürgerlichen Rechts
Az: 14-0563.1 Regierungspräsidium Karlsruhe
Vorstand: Rüdiger Dillmann, Michael Flor, Jivka Ovtcharova, Rudi Studer
Vorsitzender des Kuratoriums: Ministerialdirigent Günther Leßnerkraus