Systems and techniques for classification of incoming text messages into categories. An initial set of categories is created and populated with a set of initial messages in each category. Incoming messages are assigned to a category based on text analysis of the incoming message and analysis of dissimilarities and differences between the message and messages already in the category. A set of unclassified messages is also maintained for incoming messages that do not fit an established category. Periodically, unclassified messages are reassigned to one of the established categories or to new categories created based on analysis of the unclassified messages.