Open Source Wordpress Anomaly Detector To use this , you need to provide a path to a spam corpus, a non-spam corpus and a test corpus. To get best results use normalized data, ie data containing only words stripped of all html and headers. The output will be the probability of a test set being either spam or non-spam based upon which a classification decision can be made. There is a plan to add normalizing capabilities in addition to automating the classification going forward.