Search Mailing List Archives
[protege-owl] translating a frame-based ontology into owl
tudorache at stanford.edu
Fri Aug 31 16:33:22 PDT 2007
Olivier Dameron has a paper about the challenges involved in converting
the Foundational Model of Anatomy (FMA) from a frames representation to
OWL and some proposed solutions:
There are many subtle differences between frames and OWL, which make an
automatic translation difficult. Some of these differences were very
briefly presented here:
In Protege, there is a basic frames to OWL converter, that does a
syntactic and structural translation: the frames classes are translated
into owl:Class, the slots into datatype or object properties, the
instances into OWL individuals. The meta-classes from frames are
converted to OWL meta-classes, by this making the ontology OWL-Full. The
relationship between the frames instances are preserved in the OWL
translation, so you will be able to ask the queries you asked in frames
(the open-world vs. closed-world assumption might be a problem here).
Even though the existing translator is not complete, it might be a good
Nacho Mayorga wrote:
> I was wondering about the best or most proper way of translating an
> ontology from core-Protégé (frames) into owl. Let me explain the
> intended use case a bit:
> - the source (lightweight) ontology represents the knowledge
> underlying (every aspect of) a given software system, including both
> the system's model and meta-model (as the sw system is meant to
> reflect changes in its own model) [BTW, having a meta-model, I guess,
> prevented the automatic translation, as some fillers were classes]
> - some (typically, n-ary) relations between instances are reified
> - the target ontology should be able to answer the same queries as the
> source was able to handle (as much as possible)
> - instances in the source are to be added as individuals in the
> target, as most of the queries would refer to instances, for instance,
> + how are particular instances related to each other?,
> + what are their slot fillers?,
> + does an arbitrary predicate holds for some selected instances
> or, the other way round, which instances would satisfy a given predicate?
> The practical issues I was pondering were:
> - is there any good way of representing (and dealing with!) a
> meta-model? (such as proxy individuals representing the class as a
> singleton pattern) while still staying within owl-DL?
> - the attributes are to be represented as properties (in owl),
> aren't they?:
> + what about reified relations?:
> * how would they be best conveyed into owl?
> * would it be necessary to keep the reification or is
> there any representation pattern worth exploring?
> + what if a slot-filler was a class?
> - could the predicates in the queries (as described above) be
> represented as subsumption patterns/tests, so that a reasoner could
> check out their satisfiability?
> Help and comments would be greatly appreciated.
> Thank you very much,
> Nacho Mayorga
> protege-owl mailing list
> protege-owl at lists.stanford.edu
> Instructions for unsubscribing: http://protege.stanford.edu/doc/faq.html#01a.03
More information about the protege-owl