Ripper

Ripper is located in /proj/nlp/tools/MachineLearning/ripper/ripper
The man pages for Ripper are available here (main and auxiliary)
You can read more about ripper in Cohen '95 and Cohen '96

To run Ripper, you should have the following files:

• bla.names
  This file lists classification classes, features and the range of their values. The first line contains class names, separated by a comma. The rest of the file contains feature descriptions of the following format:
  feature1: type1.
  feature2: type2.
  In our case all the features are of type set. Eavery line in the bla.names should end with period. You can check out an example file: bla.names.
• bla.data
  This file contains your training data. Every line contains feature values for a training example separated by commas and its correct class:
  feature1, feature2, class1.
  Here is the example of bla.data.
• bla.test
  This file contains your testing data and has exactly the same format as bla.data. Here is the example of bla.test.

To run Ripper, execute the following command:
$ /proj/nlp/tools/MachineLearning/ripper/ripper bla

After training, Ripper will create file bla.hyp, which contains the classification rules.
The output of Ripper for our example is given below:


Final hypothesis is:
class2 :- right1_tag ~ NN (4/0).
default class1 (6/0).
=============================== summary ===============================
Train error rate: 0.00% +/- 0.00% (10 datapoints)
Test error rate: 0.00% +/- 0.00% (2 datapoints)
Hypothesis size: 1 rules, 2 conditions
Learning time: 0.00 sec