[Z3-zemantic] RDFS and OWL inference [was Zemantic 0.5 released]

[Olivier Grisel]

Reinout van Rees wrote:

> The OWL ontology formats are build on top of RDFS, the "simple" and
> the "description logic" restrict RDFS somewhat, making it
> calculateable. OWL full, which allows the full power of RDFS, is
> theoretically full of infinite loops.

Yes, for the OWL DL part, which is formally based on the SHOIN(Dn) 
description logic, there exists a bunch of optimized tableaux based 
algorithms that allow various inference schemes against such logics.

The interested reader should refer to the http://dl.kr.org/courses.html 
website to find introductionary and more advanced tutorials on the 
topic. I found this course especially interesting: 
http://lat.inf.tu-dresden.de/~clu/esslli.html

There also exists reference implementation such as FaCT and Racer 
although I haven't found the time toplay with them yet.

> So, the idea to allow *per-application* entailment modules to be
> plugged in sounds like a very sane idea :-)

Yes, we should keep in mind the components/adapters architecture of zope 
3 which allow a great deal of flexibility although we don't have the 
zope interfaces machinery available in pure python modules.

Best,

--
Olivier