[Top] [All Lists]

Re: [ontolog-forum] Ontologies for hybrid connectionist-semantic systems

To: ontolog-forum@xxxxxxxxxxxxxxxx
From: "John F. Sowa" <sowa@xxxxxxxxxxx>
Date: Sat, 04 Dec 2010 09:05:59 -0500
Message-id: <4CFA4AC7.1000203@xxxxxxxxxxx>
John B an Leo,    (01)

Before going further, I'd like to clarify an ambiguity, which may not
be obvious to all readers of Ontolog Forum:    (02)

  1. Following is the definition of 'connectionism' that is familiar
     to people in AI and cognitive science:    (03)

     "Connectionism is a movement in cognitive science which hopes
     to explain human intellectual abilities using artificial neural
     networks (also known as ‘neural networks’ or ‘neural nets’)."    (04)

     Source: http://plato.stanford.edu/entries/connectionism/    (05)

  2. There is an organization (cnx.org), which is developing educational
     software that promotes connections between modules of documents.
     It uses the markup language CNXML to link and relate modules.    (06)

On 12/4/2010 7:20 AM, John Bottoms wrote:
> It occurs to me that it should to possible to create an application that
> talks OWL in one direction and CnxLanguage in the other. But that might
> be like breaking your arm to scratch your back.    (07)

Unfortunately, that last sentence may be true.  Both the Semantic Web
and the CNX movement have related goals of linking documents.  For
that reason, they are on a collision course:    (08)

  1. The Semantic Web was begun in the 1990s with some ideas taken
     from the WWW and some from AI to promote a Giant Global Graph
     that spans everything on the Internet and relates it with
     semantic links.    (09)

  2. CNX is a more recent and more specialized technology for developing
     educational modules that can be processed by software.  The goal is
     to provide a unified foundation for all educational software.    (010)

The problem that John B noticed is an inevitable clash whenever two
groups are independently developing "unified foundations". Each one
has a "unified" vision of one aspect, which tends to grow until it
encroaches on the boundaries of other "unified" visions.  The result
is technically known as a "religious war" because it resembles the
clashes between "unified" visions of God and the Universe that have
been promoted and attacked by opposing zealots from time immemorial.    (011)

> While expensive, such an application might be useful in providing a
> migration path between semantic systems and AI-NLP systems.    (012)

I would like to emphasize that AI and NLP began the R & D on semantic
systems over a half century ago.  The Semantic Webbers took little
bits and pieces from that work and pasted it onto their URIs and
declared themselves to be the Universe of Semantics.    (013)

I have called the Semantic Webbers provincial.  But they protest    (014)

    "no, No, NOO! We are not provincial because we are the universe!"    (015)

But that is the definition of provincialism:  to think that your
province is the universe.    (016)

> Clearly there are attractive features of Bayesian systems that could be
> useful... if they could be integrated or designed to make use of semantic
> information and knowledge.    (017)

I strongly agree.  And many people in AI and NLP have been using
a variety of statistical methods in conjunction with other semantic
tools.  But so far, those techniques have not penetrated the province
called the Semantic Web.    (018)

And by the way, this gets to my primary complaint about OWL:  It is the
epitome of provincialism:    (019)

  1. Tim B-L had a good idea:  add semantics to the links of the WWW.    (020)

  2. Many people (including me in 1998) thought that was a great idea.    (021)

  3. Lots of people who had "unified" visions latched onto that nascent
     movement by relabeling their technology as components of the SW.    (022)

  4. There is nothing inherently wrong with point #3, but provincialism
     raised its ugly head:  Once some groups got a foothold in the SW,
     they declared their little province to be the universe, and built
     a wall around it to protect their province from outside influences.    (023)

If you read the latest OWL 2.0 documents, they have thrown in everything
plus the kitchen sink.  Instead of being one module in the SW toolkit,
OWL has grown to be another unified vision of the SW, which is itself
a unified vision of the WWW, which is itself a unified vision of IT.    (024)

Contrast that view with the way Leo has been using DL technology:    (025)

> ... when I use OWL for complex applications, I have combined it with
> SWRL and used a logic-programming environment for reasoning (Prolog
> or actually a combined DL+Prolog environment). This gives me
> approximately FOL reasoning. Or I don't use OWL and instead I use
> Cyc or HighFleet's (formerly Ontology Works) deductive database,
> or simply Prolog, or sometimes a theorem prover such as Vampire
> or Prover9. Why? Because indeed description logics are primarily
> useful for classificational reasoning. And often one wants more.    (026)

I enthusiastically support this approach.  And that is the way that
DLs were introduced in the 1980s:  DLs were designed and used as
one component of a *hybrid* system, as Leo described above.  The DL
component defined the terminology (T-Box), and a more expressive
logic (rules or full FOL) were used for assertions (A-Box).    (027)

Recommendation:  Anyone who is using OWL and finds it useful should
continue to use it as long as it is useful.  But for future projects,
OWL should be declared as a *legacy system* that must be supported,
but not a growth path for the future.    (028)

I believe that DL technology can be useful as one component of a hybrid
system.  And I would emphasize that the most important and most widely
used aspects of DLs are still the Aristotelian syllogisms.    (029)

As I pointed out, the BFO ontology does not use any feature of OWL
beyond what Aristotle defined.  I recommend that the *normative*
version of BFO be stated in a very readable Aristotelian form,
with the OWL version as just one implementation for those who may
be stuck with using OWL.    (030)

John Sowa    (031)

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    (032)

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