Knowledge Representation:

   OpenCyc -- OpenCyc is the open source version of the Cyc(r)
	      technology, the world's largest and most complete
	      general knowledge base and commonsense reasoning
	      engine. OpenCyc can be used as the basis for a wide
	      variety of intelligent applications.  

	      web site: http://www.opencyc.org
	      documentation: http://www.opencyc.org/doc
	      download: http://sourceforge.net/projects/opencyc

   KNOWBEL -- ftp://ai.toronto.edu/pub/kr/ as the files knowbel.tar.Z and
              manual.txt.tar.Z 
              Contact: Bryan M. Kramer, <kramer@ai.toronto.edu>
              Telos temporal/sorted logic system.

   SB-ONE  -- Contact: kobsa@inf-wiss.uni-konstanz.de
              KL-ONE family. Currently undergoing revision and will be
              renamed KN-PART+.
   KRIS    -- Contact: baader@dfki.uni-kl.de
              KL-ONE family (Symbolics only)
   BACK    -- Contact: back@cs.tu-berlin.de
              ftp.cs.tu-berlin.de:/pub/doc/reports/tu-berlin.de/kit/Back52
                 Files are BACK_V52.intro and Back52.tar.Z
                 Tar file includes Tutorial/Manual in postscript format
                 and installation instructions.
              KL-ONE family
   CLASSIC -- Contact: dlm@research.att.com
              KL-ONE family
   MOTEL   -- Contact: hustadt@mpi-sb.mpg.de
              ftp://mpi-sb.mpg.de/pub/tools/ [139.19.1.1]
              Modal KL-ONE (contains KRIS as a kernel). 
              Implemented in Prolog.

   FOL GETFOL -- Contact: fausto@irst.it
              Weyrauch's FOL system

   COLAB/RELFUN  -- Contact: boley@informatik.uni-kl.de
                    Logic Programming
   COLAB/FORWARD -- Contact: hinkelma@dfki.uni-kl.de
                    Logic Programming
   COLAB/CONTAX  -- Contact: meyer@dfki.uni-kl.de
                    Constraint System for Weighted Constraints over
                    Hierarchically Structured Finite Domains.
   COLAB/TAXON   -- Contact: hanschke@dfki.uni-kl.de
                    Terminological Knowl. Rep. w/Concrete Domains

   SNePS (Semantic Network Processing System) is the implementation of a
   fully intensional theory of propositional knowledge representation and
   reasoning. SNePS includes a module for creating and accessing
   propositional semantic networks, path-based inference, node-based
   inference based on SWM (a relevance logic with quantification) that
   uses natural deduction and can deal with recursive rules, forward,
   backward and bi-directional inference, nonstandard logical connectives
   and quantifiers, an assumption based TMS for belief revision (SNeBR), a
   morphological analyzer and a generalized ATN (GATN) parser for parsing
   and generating natural language, SNePSLOG, a predicate-logic-style
   interface to SNePS, XGinseng, an X-based graphics interface for
   displaying, creating and editing SNePS networks, SNACTor, a
   preliminary version of the SNePS Acting component, and SNIP 2.2, a new
   implementation of the SNePS Inference Package that uses rule shadowing
   and knowledge migration to speed up inference.  SNeRE (the SNePS
   Rational Engine), which is part of Deepak Kumar's dissertation about
   the integration of inference and acting, will replace the current
   implementation of SNACTor.  SNePS is written in Common Lisp, and has
   been tested in Allegro CL 4.1, Lucid CL 4.0, TI Common Lisp, CLISP
   May-93, and CMU CL 17b. It should also run in Symbolics CL, AKCL 1.600
   and higher, VAX Common Lisp, and MCL. The XGinseng interface is built
   on top of Garnet.  SNePS 2.1 is free according to the GNU General
   Public License version 2. The SNePS distribution is available by
   anonymous ftp from 

      ftp://ftp.cs.buffalo.edu/pub/sneps/  [128.205.32.9] 

   as the file rel-x-yyy.tar.Z, where 'x-yyy' is the version. The other
   files in the directory are included in the distribution; they are
   duplicated to let you get them without unpacking the full distribution
   if you just want the bibliography or manual. If you use SNePS, please
   send a short message to shapiro@cs.buffalo.edu and
   snwiz@cs.buffalo.edu. Please also let them know whether you'd like to
   be added to the SNUG (SNePS Users Group) mailing list. 

   URANUS is a logic-based knowledge representation language. Uranus is
   an extension of Prolog written in Common Lisp and using the syntax of
   Lisp. Uranus extends Prolog with a multiple world mechanism for
   knowledge representation and term descriptions to provide
   functional programming within the framework of logic programming.
   It is available free by anonymous ftp from 

      ftp://etlport.etl.go.jp/pub/uranus/ftp/ [192.31.197.99]

   for research purposes only.  For more information contact the author, 
   Hideyuki Nakashima <nakashim@etl.go.jp>.

Machine Learning:
 
   The prudsys XELOPES library (eXtEnded Library fOr Prudsys Embedded
   Solutions) is an open platform-independent and
   data-source-independent library for Embedded Data Mining. It was
   developed in close cooperation with the Russian MDA specialist
   ZSoft Ltd. XELOPES is CWM-compatible, supports the relevant Data
   Mining standards and can be combined with all prudsys products.
   http://www.prudsys.com/Produkte/Algorithmen/Xelopes

   RFCT is a tool based on C4.5 and written in Java. It uses C4.5 to
   discover temporal and causal rules, and has the following features:
    *) Has a graphical user interface.
    *) Handles temporal data, both in input and output.
    *) Can function in an unsupervised manne.r
    *) Outputs temporal/causal rules in a useful manner, so the user can
       have a good understanding of what influences the result.
    *) handles continous values (can discretize real-valued variables).
    *) Can output rules in Prolog, thus the rules are readily
       executable.
    The package, including full source code, example files, and online
    help, is available freely from
    http://www.cs.uregina.ca/~karimi/downloads.html.
 

   LIBSVM -- a support vector machines (SVM) library for
               classification problems by Chih-Chung Chang and
               Chih-Jen Lin. See:
               http://www.csie.ntu.edu.tw/~cjlin/libsvm

   Weka     -- a GPLed Java machine learning toolkit
	       http://www.cs.waikato.ac.nz/ml/weka/
	       Is associated with an ML book.  See:
	       http://www.cs.waikato.ac.nz/~ml/weka/book.html

   COBWEB/3 -- Contact: cobweb@ptolemy.arc.nasa.gov

   IND      -- Contact: NASA COSMIC, <service@cossack.cosmic.uga.edu>
               Tel: 706-542-3265 (ask for customer support)
               Fax: 706-542-4807
               IND is a C program for the creation and manipulation of
               decision trees from data, integrating the CART,
               ID3/C4.5, Buntine's smoothing and option trees, Wallace
               and Patrick's MML method, and Oliver and Wallace's MML
               decision graphs which extend the tree representation to
               graphs. Written by Wray Buntine, <wray@kronos.arc.nasa.gov>.

   AUTOCLASS -- Contact: taylor@ptolemy.arc.nasa.gov
               AutoClass is an unsupervised Bayesian classification system for
               independent data. 

   FOIL     -- ftp.cs.su.oz.au:/pub/ [129.78.8.208] 
               as the files foil4.sh, foil5.sh, and foil6.sh.
               Each shell archive contains source, a brief manual,
               and several sample datasets. FOIL2 should be available
               from sumex-aim.stanford.edu:/pub/FOIL.sh. FOIL 6.0 now uses
	       ANSI C.
               Contact: J. Ross Quinlan <quinlan@cs.su.oz.au>
                        Mike Cameron-Jones <mcj@cs.su.oz.au>

   RWM      -- Contact: H. Altay Guvenir <guvenir@trbilun.bitnet>
               RWM is a program for learning problem solving strategies,
               written in Common Lisp (tested on Suns and NeXT).


   MOBAL is a system for developing operational models of application
   domains in a first order logic representation. It integrates a manual
   knowledge acquisition and inspection environment, an inference engine,
   machine learning methods for automated knowledge acquisition, and a
   knowledge revision tool.  By using MOBAL's knowledge acquisition
   environment, you can incrementally develop a model of your domain in
   terms of logical facts and rules.  You can inspect the knowledge you
   have entered in text or graphics windows, augment the knowledge, or
   change it at any time. The built-in inference engine can immediately
   execute the rules you have entered to show you the consequences of
   your inputs, or answer queries about the current knowledge. MOBAL also
   builds a dynamic sort taxonomy from your inputs. If you wish, you can
   use several machine learning methods to automatically discover
   additional rules based on the facts that you have entered, or to form
   new concepts. If there are contradictions in the knowledge base due to
   incorrect rules or facts, there is a knowledge revision tool to help
   you locate the problem and fix it.  MOBAL (release 3.0b) is available
   free for non-commercial academic use by anonymous ftp from

      ftp.gmd.de:/gmd/mlt/Mobal/

   The system runs on Sun SparcStations, SunOS 4.1, and includes a
   graphical interface implemented using Tcl/TK.

   PEBLS (Parallel Exemplar-Based Learning System) is a nearest-neighbor
   learning system designed for applications where the instances have
   symbolic feature values.  PEBLS has been applied to the prediction of
   protein secondary structure and to the identification of DNA promoter
   sequences. PEBLS 3.0 is written in ANSI C and is available by
   anonymous ftp from ftp://blaze.cs.jhu.edu/pub/pebls/
   [128.220.13.50] for research purposes only. For more information,
   contact Steven Salzberg <salzberg@cs.jhu.edu>.

   OC1 (Oblique Classifier 1) is a multivariate decision tree induction
   system designed for applications where the instances have numeric
   feature values.  OC1 builds decision trees that contain linear
   combinations of one or more attributes at each internal node; these
   trees then partition the space of examples with both oblique and
   axis-parallel hyperplanes.  OC1 has been used for classification of
   data from several real world domains, such as astronomy and cancer
   diagnosis.  A technical decription of the algorithm can be found in
   the AAAI-93 paper by Sreerama K. Murthy, Simon Kasif, Steven Salzberg
   and Richard Beigel.  A postscript version of this paper is included in
   the distribution. OC1 is a written entirely in ANSI C. OC1 is
   available by anonymous ftp from  

      ftp://blaze.cs.jhu.edu/pub/oc1/ [128.220.13.50]

   This distribution is provided for non-commercial purposes only. For
   more information, contact Sreerama K. Murthy <murthy@cs.jhu.edu>
   (primary contact), Steven Salzberg <salzberg@cs.jhu.edu>, or Simon
   Kasif <kasif@cs.jhu.edu>, Department of Computer Science, The Johns
   Hopkins University, Baltimore, MD 21218.

   Set-Enumeration (SE) Trees for Induction/Classification. Significant
   research in Machine Learning, and in Statistics, has been devoted to
   the induction and use of decision trees as classifiers.  An induction
   framework which generalizes decision trees using a Set-Enumeration
   (SE) tree is outlined in  

      Rymon, R. (1993), An SE-tree-based Characterization of the Induction
      Problem. In Proc. of the Tenth International Conference on Machine
      Learning, Amherst MA, pp. 268-275.

   In this framework, called SE-Learn, rather than splitting according to
   a single attribute, one recursively branches on all (or most) relevant
   attributes. An induced SE-tree can be shown to economically embed many
   decision trees, thereby supporting a more expressive hypothesis
   representation. Also, by branching on many attributes, SE-Learn
   removes much of the algorithm-dependent search bias. Implementations
   of SE-Learn can benefit from many techniques developed for decision
   trees (e.g., attribute-selection and pruning measures). In particular,
   SE-Learn can be tailored to start off with one's favorite decision
   tree, and then improve upon it by further exploring the SE-tree. This
   hill-climbing algorithm allows trading time/space for added accuracy.
   Current studies (yet unpublished) show that SE-trees are particularly
   advantageous in domains where (relatively) few examples are available
   for training, and in noisy domains. Finally, SE-trees can provide a
   unified framework for combining induced knowledge with knowledge
   available from other sources (Rymon, 1994).  

      Rymon, R. (1994), On Kernel Rules and Prime Implicants. To appear in
      Proc. of the Twelfth National Conference on Artificial Intelligence,
      Seattle WA.

   A Lisp implementation of SE-Learn is available from Ron Rymon
   <Rymon@ISP.Pitt.edu>. A commercial version in C is currently under
   development.

   MLC++ is a Machine Learning library of C++ classes being developed at
   Stanford.  More information about the library can be obtained at URL

      /robotics.stanford.edu:/users/ronnyk/mlc.html">http://robotics.stanford.edu:/users/ronnyk/mlc.html

   The utilities are available by anonymous ftp from

      starry.stanford.edu:/pub/ronnyk/mlc/util/

   They are currently provided only as object code for Sun, but source code
   will be distributed to sites that wish to port the code to other compilers.
   For more information write to Ronny Kohavi <ronnyk@CS.Stanford.EDU>.

Medical Reasoning:

   TMYCIN -- sumex-aix.stanford.edu:/tmycin