determine spam based on a confidence rating, e.g. reply not from follower +1. reply to > 1 user, +2. contains url, +5. set a threshold ~5.