Path: utzoo!attcan!uunet!husc6!ogccse!blake!uw-beaver!cornell!batcomputer!sun.soe.clarkson.edu!rpi!liszt.cs.rpi.edu!weltyc
From: weltyc@cs.rpi.edu (Christopher A. Welty)
Newsgroups: comp.ai
Subject: Re: KEE vs other knowledge rep languages
Keywords: KEE, knowledge representation, ART, NIKL, KL-ONE
Message-ID: <312@rpi.edu>
Date: 24 Jan 89 23:45:32 GMT
References: <931@novavax.UUCP> <6140@columbia.edu>
Sender: usenet@rpi.edu
Organization: RPI Computer Science Dept.
Lines: 31

In article <6140@columbia.edu> baker@garfield.UUCP (Michelle Baker) writes:
>
>We are currently deciding on which knowledge representation language
>to use for a fairly large research project.   Currently KEE seems to
>be the candidate of choice but we are interested in comparing this to
>some of the others, e.g. ART, NIKL, HYPERCLASS, etc.  

We use CGIs Knowledge Craft at RPI, and find it much more expressive
an environment than KEE or ART.  It's not as `flashy' as some of the
other commercial knowledge tools, but it is quite powerful.  The
underlying representation language is CRL which was SRL, and it
subsumes the frame-based capabilities of ART and KEE, neither of which
I would classify as `representation languages'.

NIKL, which is based on KL-ONE, is also very expressive, but I've
never seen any applications (which by no means implies there aren't
any) that couldn't be done in KEE or ART (by this I mean couldn't be
done using the natural facilities of these frame systems).

The real key in determining which is the best for you is what
exactly you are doing.  KEE and ART are (in my humble view) far better
`production environments', good for making manager-pleasing software
with bells and whistles, but perhaps not as interesting to the
KR researcher as NIKL or Knowledgecraft.

I have lots of references on CRL, KL-ONE, and implementations using
them if anyone is interested.


Christopher Welty  ---  Asst. Director, RPI CS Labs
weltyc@cs.rpi.edu             ...!njin!nyser!weltyc