KIF: Re: SUMO axiomatization

[John F. Sowa]

Adam,

I agree that the issues have been debated ad nauseam.  I don't
believe that there is a need for any further debate, since the
proposed solution will satisfy everybody:

 1. Support both sorts and row variables in the logical foundation
    (as well as other features, such as modules).

 2. Define mappings to and from the full LF for both full standard
    KIF and full standard CGIF.

 3. Define a subset mapping for a "stripped down" KIF that contains
    only those features that you want.

AP> You
> could interpret "more complex" differently.  You could say (as Bill and
> Chris have) that "more complex" refers to the axioms that are needed to
> define particular concepts in the ontology.  They've shown good examples of
> how row variables make for more compact definitions.  Are they simpler or
> easier to understand?  It depends on what community you're addressing, and
> comes down to personal preference, not a statement of right or wrong.

There are several communities that I want to support:  

 1. The logicians who are concerned about supporting very expressive
    languages that can be used for R & D purposes in a wide range of
    areas.  I won't enumerate them here, but I'll point to a conference
    on philosophical logic, where I'll be giving a talk next May:

       http://www.philog.ruc.dk/phiconf1.html

 2. The R & D people who are developing logic-based tools for
    both theoretical purposes and commercial software.  (I include
    things like UML and SQL as examples of logic-like-based
    commercial tools that should be supported.)

 3. The researchers on natural language semantics who need
    versions of logic that can be used to support the full
    expressive power of natural languages (or as close as we
    can get to it with our best available tools and techniques).

 4. The domain experts who need easy-to-use tools that support
    languages for defining their knowledge bases and working with
    them in highly readable languages, such as controlled English
    and other NLs.  For further discussion about controlled NL and
    its use in architectures for intelligent systems, see my recent
    paper:

       http://www.jfsowa.com/pubs/arch.htm

The fourth group includes those who are most in need of an easy to
use, but expressive language.  Neither KIF nor CGIF are sufficiently
readable, but the semantic foundations we have been proposing for the
LF standard will be able to support them.  I view both KIF and CGIF
as intermediate languages, which should be used primarly like Java
bytecodes at a level between LF and the more readable languages.

You may agree or disagree with my choice of communities, but the
kind of LF that the KIF group has been developing will support both
our requirements.   There is no need for further debate about that.

John Sowa