SLIPPER is a rule-learning system. Formally, it is based on confidence-rated boosting, a variant of AdaBoost developed by Rob Schapire and Yoram Singer in 1999. The code is based on William Cohen's widely-used RIPPER learning system. A detailed description of the SLIPPER algorithm can be found in a technical paper by Cohen and Singer in AAAI-99, titled A Simple, Fast, and Effective Rule Learner.
Like RIPPER, SLIPPER is fast, robust, and easy to use, and its hypotheses are compact and easy to understand. Like RIPPER, SLIPPER supports set-valued features, which makes it useful for text categorization using a "bag of words" representation of text. Unlike RIPPER, SLIPPER is unencumbered by patent constraints, and it is freely available for research purposes.
A version of SLIPPER suitable for academic and research use can be downloaded from Rutgers University. (Thanks to Haym Hirsh for making this legally possible.) It is available as C source code, compiling under the GNU C compiler, or as executables for Linux or Windows. (For Windows, Cygwin must be installed. Cygwin is a free package of Unix-compatible tools and libraries.)
There is also a tarball of sample datasets in SLIPPER format available (about 300Kb). The data in the "pdata" subdirectory is propositional, and the data in the "tdata" subdirectory is text.
Information about commercial licensing is also available on request.
For licensing information or support questions, contact William Cohen at firstname.lastname@example.org