[Top] [All Lists]

Re: [ontolog-forum] Truth

To: "[ontolog-forum] " <ontolog-forum@xxxxxxxxxxxxxxxx>
From: Chris Mungall <cjmungall@xxxxxxx>
Date: Mon, 9 Jul 2012 22:58:15 -0700
Message-id: <469E7795-DB94-4A54-90AD-9C3D77613761@xxxxxxx>

On Jul 9, 2012, at 8:05 PM, John F Sowa wrote:    (01)

> Leo and Kingsley,
> I'm responding to both of your notes because they address two aspects
> of a disease that I believe has been destroying the Semantic Web.
> In short, the Semantic Web is first and foremost an engineering problem,
> but it has been infected by certain scientists who forced the engineers
> to adopt inappropriate solutions to their problems -- or even worse,
> to ignore certain important kinds of problems.
> Leo
>> Motik, Boris. 2005. On the Properties of Metamodeling in OWL.
>> In: 4th Int. Semantic Web Conf. (ISWC 2005).
> I attended a talk by Boris Motik, and I spoke with him.  He is very good
> at what he does.  This paper demonstrates his technical ability.  But
> like nearly all of the DL theoreticians, he knows nothing about how to
> develop applications or what practitioners would ever find useful.    (02)

Boris Motik developed HermiT, which I would consider a very useful application.    (03)

For the Gene Ontology (a major application in molecular biology) we rely on 
HermiT to verify
every change made to the ontology. This counts as a useful application to us.    (04)

Boris also developed KAON2 and its underlying datalog engine, which I'm less 
familiar with.    (05)

> He is very good at proving theorems about decidability, but nobody has
> shown that restricting a language to make it decidable is useful for
> any purpose whatsoever.  Restricting a language cannot solve anything
> faster.  It only makes certain kinds of problems impossible to state.    (06)

I agree that decidability has unduly elevated by the DL community, and
that hardwiring the restriction into the language is limiting. Ideally, OWL2-DL 
would be 
standardized as profile of Common Logic, just as OWL-EL++ is a profile of DL. 
Modelers should use what is appropriate for representing the domain, and then 
reasoners based on individual requirements. In many cases, the EL++ subset
of DL is sufficient, and there are some very fast reasoners for this subset 
the lightning-fast Elk, also developed in Ian Horrocks' group).    (07)

But the fact that decidability has been elevated does not invalidate the useful 
algorithms and applications that have been developed by this community/    (08)

> Cyc has had far more experience in working on actual problems than
> any of the DL theoreticians ever dreamed of.    (09)

I am aware of many DL theoreticians that have worked closely with developers
of biomedical ontologies resulting in applications that have been widely used, 
actual problems.    (010)

(I'm not aware of problems actually solved by Cyc, but this is outside my area 
of expertise)    (011)

>  And they have *never*
> found decidability to be a problem.  Bob MacGregor developed the widely
> used LOOM and PowerLoom systems, which combined a DL with a rule-based
> system and with bindings to programming languages and databases.    (012)

I like LOOM, I will have to give it another try.    (013)

> MacGregor worked with users who actually used his systems to develop
> major applications.  And he said that *none* of the users ever asked
> for decidability, but they all asked for more expressive power.
> The net result is that the "Decidability Thought Police" purged
> MacGregor from their community.
> Decidability is the single worst disease that has destroyed the
> usefulness of the SW.  When I look at Motik's paper, I see disease.
> There is nothing in that paper that is of the slightest value for
> any practical application of any kind.  It makes me angry that
> those people have been destroying what might have been a very
> useful development, if they had left the SW to the engineers.    (014)

I think that's a bit much.    (015)

> JFS    (016)

Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/  
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/ 
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J    (017)

<Prev in Thread] Current Thread [Next in Thread>