[seek-kr-sms] ontology management

[Joseph Goguen]

Hi Shawn, thanks for all this; it's very helpful.

Shawn Bowers wrote:

>
> (I think) the notion of extensions makes sense to me now.  I wasn't
> quite parsing your definition correctly earlier.  In particular, as I
> understand it now, an extension maps a "core" theory into another
> theory. When you say 'theory' you mean a formal theory, but in a
> general sense. For example, C might be expressed as a first-order
> theory (but could also be represeted in some other formal system).

Ontologies in OWL (in any of its versions) are formal logical theories
since the OWL specs give a formal logical semantics; the same goes
for all the DLs and most recent ontology langauges (but not EML).

>
> The other thing that didn't parse for me, but now I understand is when
> you say "e_i: C -> C_i of C".  Here you mean that the set of
> extensions e_i are of C (i.e., the set of extensions E s.t. C is the
> domain of each e_i \in E), and I was totally confused because I
> thought you were saying the C_i (the co-domain) was "of" C.

OK, i just changed the wording.  But E is not the "set of all" but "some
given set of" maps (which i guess you know).

>
> I think you might consider adding some (simple, but) concrete examples
> of extensions (or examples of sets of extensions). And maybe even show
> how different sequences of extensions lead to an "evolution trace".
> It isn't clear to me how storing the extensions (mappings) would be
> used, i.e., what tools you have to exploit the extensions other than
> knowing that one theory came from another (because it is the co-domain
> of the mapping).  Examples of how one would exploit these extensions
> in a cvs system would als be useful, or other benefits of the approach
> such as how it could reduce storage size, etc.  I'm sure you've
> thought about this stuff, I just don't have a good feel for how it
> might be leveraged.

Yes, you are right, i should add all that..... if i can find time.  
Meanwhile
please see my just sent message to Serguei.  One good reason for saving
maps (in consistent environments of maps and theories with CVS) is to
support incremental updating of maps as the theories evolve; one doesnt
want to have to start from scratch, but one does want to start from the
right version for one's particular situation.  And as i said to Serguei, 
such
maps can translate and integrate data and concepts.  In addition, and i
think this would be a really exciting thing to explore, they can help domain
scientists find areas where they almost agree and could construct further
translations, and where they might agree if further research were done
(this is sketched in the note).

    - joseph

>
> Thanks,
> -shawn
>
>
>
>
>
> Joseph Goguen wrote:
> > Ive converted my note to html (so subscripts work), added some details
> > and clarifications (e.g., about extension maps), and put it on the 
> web, at
> >
> >    http://www.cs.ucsd.edu/~goguen/papers/onto-intgn.html
> >
> > Further omments are very welcome,
> >
> >    joseph
> >
> > Shawn Bowers wrote:
> >
> >>
> >> Hi Joseph,
> >>
> >> The ideas you discuss in this write-up I think are right on track.
> >>
> >> In some sense I've been pushing this notion of a "kernel" for a while
> >> for seek ... and I think that the notion of measurement and the
> >> surrounding framework is precisely what that "kernel" should consist
> >> of.
> >>
> >> I have a few points and questions about the write-up.
> >>
> >> First, I'm not an ontologist, but when you say "Many of them seem to
> >> believe in the possiblity of a single unified ontology that attracts
> >> consensus ...", I wonder if this is really the case. In particular, I
> >> think that *most* (but perhaps not all) people agree that it is not
> >> possible to define *the* single unified ontology. It is possible to
> >> define ontologies with broad scopes, as well as ontologies with narrow
> >> scopes, but these may not be accepted by anyone as *the* ontology for
> >> a domain.  However, the reason folks consider creating such
> >> ontologies, is because even if they are not *the* ontology, they still
> >> permit something that people can commit to ... like a formal glossary
> >> of terms.  An ontology, as you say, is just some theory, that may or
> >> may not (typically not) be completely accurate. But, still can have
> >> value (and some shimmer of "truth").
> >>
> >> In terms of calling an ontology a theory, I think you are in good
> >> company: Guarino, Wand, and particularly Bunge view ontologies in
> >> exactly this way -- in fact, to Bunge these are synonyms.  In some of
> >> his writing, Bunge uses the phrase "asking ontological questions,"
> >> which are essentially questions that "probe" the theory, to see if
> >> some fact follows from it (i.e., is entailed, or can be explained by
> >> the theory), or even to test the theory (sort of QA/QC kind of
> >> process).
> >>
> >> You say: "Such tools can also be used to identify subdomains where
> >> consensus is most likely to be achievable ..."  This argues for some
> >> mechanism to rank or denote when some fragment of an ontology is more
> >> "authoritative" than other parts. I think this notion of
> >> "authoritativeness" can be (will also be) a crucial aspect in using an
> >> ontology for reasoning/inference, e.g., in data integration.  It can
> >> provide some richer context/guidance for applying certain integration
> >> strategies, or ranking different possible strategies.
> >>
> >> I am not sure I understand the notion/definition of "extension." In
> >> particular, it looks as though given a kernel C of concepts, that the
> >> extension operator maps concepts of C to concepts of C (i.e., it maps
> >> concepts of the kernel back into the kernel). (It wasn't clear what
> >> the "of" meant in C_i of C.) I would have expected that somehow the
> >> kernel is extended by adding new concepts, related those concepts, and
> >> possibly at some point in the future either "normalizing" them (i.e.,
> >> realizing that they map to existing kernel notions), adding them to
> >> the kernel, or dropping them as being "junk".  Also, how would one
> >> handle properties or characterstics of concepts (i.e., "extend" or
> >> "modify" concepts at a finer granularity).
> >>
> >> Thanks Joseph for sending out this draft.
> >>
> >> -shawn
> >>
> >>
> >>
> >>
> >>
> >> Joseph Goguen wrote:
> >> > Dear Shawn,
> >> >
> >> > At the KR/SMS section of the SEEK AHM, i made some suggestions about
> >> > this, which i subsequently wrote up and circulated.  Just now, 
> ive put
> >> > it on my
> >> > website, at
> >> >
> >> >    http://www.cs.ucsd.edu/~goguen/papers/onto-intgn.txt
> >> >
> >> > It starts off a bit philosophical but i think gets quite 
> practical by
> >> > the end,
> >> > and also mentions the supporting theory; i still need to add 
> citations
> >> > though....
> >> >
> >> > We all missed you at the meeting but admired all the work that 
> you have
> >> > done.
> >> >
> >> > Cheers,
> >> >
> >> >    joseph
> >> >
> >> > Shawn Bowers wrote:
> >> >
> >> >> Hi all,
> >> >>
> >> >> Since I wasn't at the AHM, I'm not sure if any discussion or 
> progress
> >> >> was made in terms of Kepler/SEEK strategies and infrastructure for
> >> >> managing ontologies.
> >> >>
> >> >> Recently, KOAN2 was released with an impressive list of 
> features.  I
> >> >> wonder if this is something that we should look at more carefully,
> >> >> and possibly adopt for Kepler/SEEK.
> >> >>
> >> >> http://kaon2.semanticweb.org/
> >> >>
> >> >>
> >> >> -shawn
> >> >>
> >> >>
> >> >> _______________________________________________
> >> >> Seek-kr-sms mailing list
> >> >> Seek-kr-sms at ecoinformatics.org
> >> >>
> >> 
> http://mercury.nceas.ucsb.edu/ecoinformatics/mailman/listinfo/seek-kr-sms
> >> >>
> >> >>
> >>
>