Relocating this conversation to [ontolog-forum] ... (01)
=ppy
-- (02)
---------- Forwarded message ----------
From: Ian Horrocks <ian.horrocks@xxxxxxxxxxxxxxx>
Date: Thu, Dec 9, 2010 at 3:32 PM
Subject: Re: [ontology-summit] Invitation to a brainstorming call for the 2011
Ontology Summit
To: Ontology Summit 2011 discussion <ontology-summit@xxxxxxxxxxxxxxxx> (03)
On 9 Dec 2010, at 07:21, Bradley Shoebottom wrote: (04)
> John,
>
> Our TopQuadrant part of the trial used Sesame. The KIM part of our Trial used
>Ontotext's BigOWLIM.
>
> Working with Ontotext was my first exposure to SeRQL. I have not yet done a
>side by side comparison to SPARQL. SPARQL we like because it is a standard,
>and when mixed with SPIN and SPARQLMotion, can do some interesting functions,
>referencing to external script processing, and reasoning tasks.
>
> OWL-DL was the most suitable for what we wanted to do with axioms.
>
> I attended a recent TopQuadrant training session in which it was expressed
>that OWL-2 was overly complex and that OWL-RL could do everything that OWL 2
>could with less restrictions. (05)
I find this statement to be strange tending to disingenuous. OWL 2 RL defines a
subset of OWL 2 such that reasoning via forward chaining materialisation (a
technique that is commonly used in DB and triple-store based implementations)
is complete for atomic queries (roughly speaking). OWL 2 RL includes more or
less all of the syntactic constructors that occur in OWL 2, so in this sense it
is equally complex. However, in order to enjoy the above mentioned completeness
guarantee, it is necessary to restrict the way that constructors can be used,
with some only being allowed on the left hand side of axioms (i.e., in the
antecedent) and some only being allowed on the right hand side (i.e., the
consequent). To my way of thinking, this makes OWL 2 RL *more* complex, because
in addition to understanding the constructors, users also have to understand
constraints on their applicability that, from their POV, probably seem
strange/arbitrary. (06)
In practice, many OWL 2 RL implementations don't apply the above mentioned
restrictions, and will accept any OWL 2 ontology. However, in this case,
reasoning will no longer be complete in general, and can even be unsound if
failure to answer "yes" is treated as a "no" answer (as, in my experience, it
often is). This can surely only make things more complex from a user POV, as
they may now need to be aware of this "idiosyncratic" behaviour, and use
caution in the way they interpret query answers. (07)
The reason for defining OWL 2 RL in the first place is that forward chaining
materialisation works well with DBs and triple-stores, and allows implementors
to take advantage of their scalable query answering over large data sets. In
fact, it can be shown that query answering in OWL 2 RL is possible in time that
in the worst case increases only polynomially with the size of the data. In
*this* sense, OWL 2 RL really is less (computationally) complex. However, as I
mentioned above, the price users pay for this is an *increase* in syntactic or
cognitive complexity. (08)
Regards,
Ian (09)
> Since I do not have formal training in Description Logic (hoping to fix that
>problem this year), I cannot comment in detail.
>
> Bradley Shoebottom
> Information Architect - R&D, Innovatia Inc.
> Tel: (506) 674-5439 | Skype: bradleyshoebottom | Toll-Free: 1-800-363-335
>begin_of_the_skype_highlighting
> bradley.shoebottom@xxxxxxxxxxxxx | www.innovatia.net | Follow us on Twitter (010)
> -----Original Message-----
> From: ontology-summit-bounces@xxxxxxxxxxxxxxxx
>[mailto:ontology-summit-bounces@xxxxxxxxxxxxxxxx] On Behalf Of John F. Sowa
> Sent: December-09-10 11:04 AM
> To: ontology-summit@xxxxxxxxxxxxxxxx
> Subject: Re: [ontology-summit] Invitation to a brainstorming call for the
>2011 Ontology Summit
>
> Dear Matthew and Peter,
>
> MW:
>> ... my forthcoming book "Developing High Quality Data Models". Substitute
>> ontology for data model and the same argument applies. The benefits come
>> from improving and automating decision making through fit-for-purpose
>> information to support those decisions.
>
> I very strongly agree. Software engineers have been doing ontology
> (avant la lettre, as they say) for a very long time. And much of that
> work has been very good -- sometimes much better than what people are
> doing with so-called ontology languages.
>
> PY:
>> Most importantly, please made sure you [Bradley] are in a position to openly
>share
>> the content or technology you are contributing (ref. our Open IPR Policy...
>
> I agree that the Open IPR policy is important, but I noticed that
> Bradley's slides showed that they have also been using the Sesame
> technology by the OpenRDF organization:
>
> http://www.openrdf.org/related.jsp
>
> I don't know what other technology Bradley & Co. are using, but there
> are some interesting features of Sesame that go beyond the current
> W3C recommendations.
>
> In particular, their query language SerQL seems to be much nicer
> than SPARQL. I haven't used that technology, and I'd like to ask
> anybody who has what they think about it.
>
> This also raises another question about OWL, which has dominated
> many of the discussions. I certainly admit that OWL has been
> used for many ontologies, but many of them (BFO, for example)
> use only the most trivial aspects -- which don't go beyond what
> Aristotle specified two and a half millennia ago.
>
> There is also much more to ontology than OWL, and many people,
> such as Adam Pease, have been extremely negative about it.
> (I'm always happy to have somebody make my criticisms sound
> moderate, by comparison.)
>
> My three major criticisms about OWL are (a) it's too limited,
> (b) it's too big, and (c) it keeps growing.
>
> One might think that (a) and (b) are contradictory, but my point
> is that the type hierarchy is only one component. The original
> DLs from the 1970s and '80s were designed to be parts of a hybrid
> system, not the whole reasoning system. In fact, that is how
> Leo said they use OWL at Mitre.
>
> Using a DL as part of a hybrid system is a very flexible and
> open-ended way of designing a system. No matter how many features
> are thrown into the OWL pot, it is deliberately designed to be
> highly restricted. That means it can *never* become the whole
> system. Most applications that use some version of OWL also use
> other kinds of reasoning.
>
> But the OWL developers have almost completely dominated the logic
> part of the Semantic Web. There is a box labeled RIF, but the
> OWL developers have included their own rule-like notation for
> the (still very limited) OWL semantics. That seems like a ploy
> to strangle the competition before it can get off the ground.
>
> I keep mentioning Cyc because it shows that there is a better
> way: you don't need to restrict the user's language in order
> to support DL-style reasoning. Instead, you can allow the
> knowledge engineers or subject matter experts to state what
> they know in whatever form they find convenient.
>
> Then for any particular problem or question, the system can
> *automatically* select whichever reasoning method is appropriate
> for that task. If a DL reasoning style is sufficient, Cyc uses
> it. But it also has a wide range of other options.
>
> At our company VivoMind, we have a different approach to support
> a wide range of formal and informal methods. But the point is
> that we can use DL-style reasoning, when appropriate, without
> forcing the users to abide by a tightly constrained notation.
>
> In short, the OWL mindset is putting a stranglehold on ontology.
> But ontologies can be, should be, and have been used in a much
> wider range from Aristotle's syllogisms to Cyc and to open
> ended mixtures of formal and informal methods.
>
> I don't believe that we should deprecate OWL, but we have to
> educate people that ontology is much, much bigger (and often
> much, much simpler) than OWL.
>
> John
>
> _________________________________________________________________
> Msg Archives: http://ontolog.cim3.net/forum/ontology-summit/
> Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontology-summit/
> Unsubscribe: mailto:ontology-summit-leave@xxxxxxxxxxxxxxxx
> Community Files: http://ontolog.cim3.net/file/work/OntologySummit2010/
> Community Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?OntologySummit2010
> Community Portal: http://ontolog.cim3.net/wiki/ (011)
_________________________________________________________________
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
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx (012)
|