Comparing content-filter techniques for stopping spam

There are many new theoretical techniques for detecting spam e-mail based upon the message contents. Although Bayesian methods are the most wellknown, there are other approaches for classifying information. This paper establishes some criteria for measuring spam filter effectiveness and compares the Boosting and Support Vector Machine approaches with some well-known existing filter software. It also examines ways of transforming e-mail messages into a form which is more readily processable by such algorithms