The document discusses two papers about learning rules and classifiers from text documents: 1) The first paper evaluates using keyword-spotting rules for classifying emails in terms of accuracy and runtime compared to TF-IDF weighting. It finds that rules perform well when categories are semantically defined but provides no examples of learned rules. 2) The second paper explores using decision tables as a simple hypothesis space for classification and presents an algorithm for inducing decision tables that searches the feature space efficiently. It finds decision tables can achieve high accuracy, especially for discrete features.