Semi-automatic stemmer evaluation

In this paper two stemmers which extract stemming rules from a sample dictionary of transformations are compared. We present their capability to generalize the information in the dictionary. The factors which affect such a generalization are also shown and discussed.