David Reynolds wrote:
Have you tried using spamassassin and sa-learn to train it what is spam?
Unless I'm missing something (which is quite possible) this won't help with the image spam that I'm trying to work on, unless I can convert the image text to real text.