Abstract: A problem is presented to classify unlabeled text documents without training set.Class associated words are the words which represent the subject of classes and provide prior knowledge for training a classifier.A learning algorithm，based on the combination of Expectation-Maximization（EM） and a Naïve Bayes classifier，is introduced to classify documents from fully unlabeled documents using class associated words.In the algorithm，class associated words are used to set classification constraints during learning process to restrict to classify documents into corresponding class labels and improve the classification accuracy. Experiment results show that the technique can solve the problem with much high accuracy，and the classification accuracy with constraints is higher than that without constraints.