[Top] [All Lists]

[ontolog-forum] IKL, NL semantics, and many-sorted logic

To: Pat Hayes <phayes@xxxxxxx>, rhm@xxxxxxxxxxxxx, "Richard H. McCullough" <rhmccullough@xxxxxxxxx>, "'[ontolog-forum] '" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: John F Sowa <sowa@xxxxxxxxxxx>
Date: Sun, 30 Mar 2014 11:36:44 -0400
Message-id: <53383A0C.9000705@xxxxxxxxxxx>
Pat and Dick,    (01)

>> Is there any new IKL documentation in the last 5 years?    (02)

> No, the project was never developed once the funding had finished.    (03)

The IKRIS project from 2005 to 2006 included an impressive group of AI
experts.  Unfortunately, the ideas and results were never published,
and the organizers at MITRE deleted the reports from their web site.    (04)

For IKL spec's:  http://www.ihmc.us/users/phayes/IKL/SPEC/SPEC.htm    (05)

For Users' Guide:  http://www.ihmc.us/users/phayes/IKL/GUIDE/GUIDE.htm    (06)

I posted what I had saved and what I found on the WWW on my web site:
http://www.jfsowa.com/ikl/ .  In particular, I recommend the report
of the Evaluation Working Group, which demonstrated the use of IKL
as an interchange format among three highly expressive logics:    (07)

  * Common Logic (CL) used in KANI
  * Multi-Sorted Logic (MSL) used in Slate
  * CycL used in the Cyc knowledge base and the Noscape system    (08)

Doug Lenat said that neither KIF nor CLIF could represent all the
features of CycL, but CL with the IKL extensions could.    (09)

More recently, I wrote a brief summary of the use of the IKL
features to support metalanguage and natural language semantics.
See pp. 17 to 23 of http://www.jfsowa.com/pubs/eg2cg.pdf    (010)

For issues about many-sorted logic (MSL), see the references and
quotations in my "Notes on MSL":  http://www.jfsowa.com/logic/sorts.pdf    (011)

That article cites several articles and slides by Hassan At-Kaci.
His LOGIN system (1986) combines a hierarchy of sorts on top of
Prolog.  That combination provides a *single language* that is
faster than OWL for reasoning about taxonomies and one of the
fastest versions of Prolog for rule-based reasoning.    (012)

His more recent (2013) CEDAR system can import OWL ontologies and
run circles around OWL in performance.  Following is a summary:    (013)

Page 2 of sorts.pdf:
> Amir and At-Kaci (2013) compared the CEDAR system, which uses
> an order-sorted logic (OSL) for classifying and querying very large
> taxonomies, to six OWL-based reasoners:  Fact++, HermiT, Pellet,
> TrOWL, RacerPro, and SnoRocket. They compared them on four taxonomies
> that ranged in size from 111,559 sorts or classes (Wikipedia) to
> 903,617 sorts (NCBI). For classification, CEDAR was among the three
> fastest for all the taxonomies; on the Wikipedia taxonomy, it was
> five times faster than the second best (Fact++). For querying,
> CEDAR beat all the others by several orders of magnitude. The query
> time is the most important, since a classified CEDAR taxonomy can be
> saved and reused. CEDAR also detects cycles in the taxonomy, which
> are a serious source of inconsistencies.  For related issues, see
> the three slide presentations by At-Kaci (2013).    (014)

Note that CEDAR achieves that performance with a *more expressive*
language than OWL -- and it outperforms OWL by importing very large
OWL ontologies and running the same queries.    (015)

Those people who say that you need to restrict expressive power in order
to achieve performance are partly right -- in the sense that different
*algorithms* are designed for different subsets of logic.    (016)

But that does not -- most definitely *not* -- mean that the user should
use a restricted language to get that performance.  CEDAR and many other
systems have multiple algorithms under the cover, and they use them in
combination on different aspects of any particular problem.    (017)

For further discussion of these issues, see
    Fads and fallacies about logic    (018)

Fundamental point:  If anybody says that it's necessary to restrict a
logic to improve performance, they are confusing two distinct issues:
the language the user sees and the algorithms that support it.    (019)

Please enlighten such people by sending them a copy of this note.    (020)

John    (021)

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    (022)

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