[Top] [All Lists]

Re: [ontolog-forum] Model Semantics, Representation Syntax, and Systems

To: ontolog-forum@xxxxxxxxxxxxxxxx
Cc: "architecture-ecosystem@xxxxxxx" <architecture-ecosystem@xxxxxxx>
From: "John F. Sowa" <sowa@xxxxxxxxxxx>
Date: Thu, 11 Nov 2010 17:53:40 -0500
Message-id: <4CDC73F4.4030400@xxxxxxxxxxx>
Kingley, Alan, Doug, Chris, et al.,    (01)

> ... maybe we could use this thread to arrive at obvious  common
> ground re. data integration and the diminishing need for  a
> syntax level lingua franca.    (02)

Many philosophers, politicians, computer programmers, and even
scientists have "one factor" theories, AKA "silver bullets."
They have a single magic slogan or principle, which they claim
solves everything or at least everything they think is important.
A lingua franca is useful, but it must be sufficiently broad
to support all the paradigms anybody requires.    (03)

> I think there *is* a need for a lingua franca for intercomputer
> communication. But I support the idea that there should be
> alternative syntaxes (as long as they can be clearly translated to
> the lingua franca).    (04)

That's fine.  I have supported Common Logic because it is broad
enough to encompass a very wide range of popular methods and systems.
But even then I believe it is important to have more expressive power
than CL.  That's why I would recommend the IKL extensions for the
next version of the CL standard.  (And for research purposes, I would
encourage even more general studies.)    (05)

> I haven't checked on CL, but IMO, it should have a normative
> syntax that is considered the one that projects should make  sure
> they can produce, so as to remove the aforementioned barrier.    (06)

The CL standard is defined in terms of an abstract syntax.  As Chris
said, the CLIF concrete syntax is sufficiently similar to KIF that it
is a de facto starting point for many projects.    (07)

> The main practical issue with using unrestricted CL is that there
> are few systems that can reason (in a predictable way) over it.    (08)

That misses the point.  A restricted notation can only guarantee 
predictable reasoning is for a very narrow range of problems.
That is "magic bullet" thinking.  Any notation optimized for one
narrow range is guaranteed to be useless for infinitely many
equally worthy problems.    (09)

Doug F. worked on the Cyc project, whose CycL language has the
expressive power of IKL, and Cyc has developed methodologies for
solving a wide range of problems in a predictable way within an
expressive framework.  It supports a family of reasoning methods
for different kinds of problems under a very broad umbrella.  I
endorse Doug's response:    (010)

> Where is the problem here?  An interlingua must be at least as powerful
> as the languages between which it is used to translate.  Although
> knowledge bases which it is used to translate may not exercise all the
> capabilities of the interlingua, the interlingua could use higher order
> expressions.  The systems which translate to and from the interlingua
> would be designed to do just that, and not act as generic theorem
> provers.    (011)

At a somewhat less expressive level than Cyc, but with a much larger
base of practical implementations, I would cite the UML diagrams,
each of which expresses a different subset of FOL.  The major flaw
in the original version of UML is that they did not take the obvious
next step of using the common foundation (FOL) to map between the
different diagrams.  More recently, they have mapped the UML diagrams
to Common Logic, but they haven't yet integrated those mappings with
their design methodologies.    (012)

> One language that DOES have these features is one that has not been used
> for the Semantic Web because its native syntax is not RDF: CycL.  CycL
> not only could be used to express mappings among different ontologies,
> but since it has its own massive ontology, hundreds of thousands of the
> ontology terms would already be expressed in the language.    (013)

I agree.  But Lenat & Co. admitted that the IKL extensions to CL have
the same expressive power as CycL.  In fact, the IKRIS project showed
that IKL could be used for interoperability in communications among
several different AI systems, including Cyc.    (014)

At the end of this note is a slide from a talk I presented at the RuleML
conference in October.  It summarizes the work of the IKRIS project for
demonstrating interoperability.  See the full set of slides for details.    (015)

> Common Logic... is a semantic framework that supports an unlimited
> number of alternative languages -- although it does not privilege
> any particular language (a.k.a., CL dialect) over any other
> (although the KIF-like dialect CLIF is sort of a default).    (016)

I agree, but I wouldn't claim that CL or even IKL, by itself, is
a magic bullet that can solve all problems.  The full range of
problems is enormous, and the following slides are an attempt to
show the magnitude of the issues and some ways to address them:    (017)

    Integrating Semantic Systems    (018)

Summary:  You can't support interoperability among all IT systems
by narrowing the expressive power.  Users always ask for *more*,
not less expressive power.  Cyc and UML have shown how to avoid
getting trapped in a single-paradigm, magic-bullet approach:
use methodologies with an open-ended variety of design patterns
that can guarantee efficiency on different classes of problems.    (019)

____________________________________________________________________    (020)

Source:  http://www.jfsowa.com/talks/clruleml.pdf    (021)

IKRIS Project    (022)

DoD-sponsored project: Design an Interoperable Knowledge
Language (IKL) as an extension to Common Logic.    (023)

Goals:    (024)

   ● Enable interoperability among advanced reasoning systems.
   ● Test that capability on highly expressive AI languages.    (025)

Show that semantics is preserved in round-trip mapping tests:    (026)

   ● Cycorp: Cyc Language → IKL → CycL
   ● RPI / Booz-Allen: Multi-Sorted Logic → IKL → MSL
   ● Stanford/IBM/Battelle: KIF → IKL → KIF
   ● KIF → IKL → CycL → IKL → MSL → IKL → KIF    (027)

Conclusion: “IKRIS protocols and translation technologies
function as planned for the sample problems addressed.”    (028)

Interoperable Knowledge Representation for Intelligence Support
(IKRIS), Evaluation Working Group Report, prepared by David A. Thurman,
Alan R. Chappell, and Chris Welty, Mitre Public Release Case #07-1111.    (029)

http://nrrc.mitre.org/NRRC/Docs_Data/ikris/IKRIS_Evaluation_Report_31Dec06.doc    (030)

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    (031)

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