In other answers at Stackoverflow it's been suggested that Weka is good, but there are others (Classifier4j, jBNC, Naiban).
Does anyone have actual experience with these?
Thomi
Weka is awesome, but Classifier4J is probably closer to what you want, since it's more aimed at text recognition.
You could also have a look at the still very new Apache Mahout project.
Another new library is ci-bayes, nice and very simple to use.
disclaimer: I have submitted a patches to this project and am currently a committer
来源：https://stackoverflow.com/questions/480683/whats-the-best-open-source-java-bayesian-spam-filter-library