W3C home > Mailing lists > Public > public-owl-dev@w3.org > October to December 2007

RE: Inferring Properties based on Types

From: Michael Schneider <schneid@fzi.de>
Date: Tue, 9 Oct 2007 11:13:54 +0200
Message-ID: <0EF30CAA69519C4CB91D01481AEA06A040A829@judith.fzi.de>
To: "John McClure" <jmcclure@hypergrove.com>
Cc: "Swanson, Tim" <tim.swanson@semanticarts.com>, "Owl Dev" <public-owl-dev@w3.org>


I can see that Tim has already constructed a nice example which demonstrates
the problem. My own example would have been a little closer to the original
problem, having an upper property 'grandParent' (as a substitute for
'sibling'), and two sub properties 'grandFatherViaFather' and
'grandFatherViaMother', which both happen to have range 'HumanMale', but
which will in real life be considered not to be equivalent. (This
distinction is btw. explictly done by Chinese people.) Ok, not necessary to
explain this anymore, I can instead comment on another important point.

John McClure wrote:


>Maybe I don't understand the original question, but it seems 
>the answers
>provided invoke  'workarounds' with Property chains and 
>what-not, are pretty
>shocking requirement when expressing such a common design 
>pattern.  Maybe the
>issue here is somewhat purity vs practicality?

John, please note that this whole discussion was about the specific
question, if a certain rule scheme is expressible in OWL or not, i.e. if
there is a set of OWL axioms, by which this rule scheme can be simulated.
OWL has not been designed to work as a rule language, and so it should not
come to a big surprise to you that some rule schemes happen to be difficult
or impossible to simulate in OWL, even if the rule scheme looks pretty easy
as in Tim's case. Generally, it will really be of /practical/ relevance to
enable people to define certain types of relationships between classes and
properties in the form of rules, while these rules can then be translated by
the OWL reasoner (in a pre processing step) into an equivalent set of
axioms. Knowing such a transformation, as I found it in one of my previous
mails, does *not* mean that *you* have to write down the scary looking
axioms. It means instead that (in principle) *you* can write down the
convenient rule, while it will be the *reasoner's* task to silently perform
the transformation into the set of axioms.

However, this whole discussion was *not* about the question if it will be
possible or not, or hard, to apply such a rule in the *Semantic Web*. If
*this* was your concern, then I can tell you: Don't panic! :-) In many
practical situations the following approach will certainly meet most
people's needs: Simply publish your example ontology, containing these RDFS
axioms like "brother ISA sibling", etc., and also publish your actual data,
like "Paul is a sibling of John". Then, in the easiest case, you can simply
add a SPARQL endpoint on top of this data, and let people put the following

  construct { :John :brother ?b }
  where { 
    :John :sibling ?b . 
    ?b a :HumanMale .

This will extract "all" brothers of :John, at least all those which can be
extracted directly from your data (not those living in the remaining "open
world" of the Semantic Web). Of course, this will only work that easy, if
you have defined all 'sibling' relationships and 'HumanMale' classifications
explicitly. If you have not, using an RDFS triple store like Sesame might
perhaps already suffice to help you in inferring the missing 'HumanMale'
classification. And if you did not want to specify all 'sibling'
relationships explicitly (maybe John is part of a large family), then you
can use, for example, an OWL reasoner like Pellet in cooperation with Jena,
to compute all 'sibling' relationships by evaluating additional transitivity
and symmetry axioms for ':sibling'. In the latter case, the SPARQL query
would then be performed on an "InferenceGraph", which is build on top of the
original assertions. Or you can leave the original assertions as they are,
and compute the missing 'sibling' relationships with your own specialised
rule engine on top of the assertions. Jena, for example, provides such a
means, and this might in some cases be more efficient compared to using a
full featured OWL/DL reasoner like Pellet.  

Anyway, you see that this problem can be solved in a variaty of more or less
efficient and hopefully less scary looking ways in the Semantic Web. The
success of the Semantic Web will certainly not completely depend on the
question, if Tim's 'brother' rule can be translated into OWL axioms or not.


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
Received on Tuesday, 9 October 2007 09:14:16 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:58:15 UTC