jezhumble

• Bayesian filter to detect potential abusers, filtering on words, length of time the user has been active, number of tweets a user has sent, who the user has tweeted to, # of spam reports against the user
• Anyone who is tweeted to on a regular basis by potential abusers gets flagged (not publicly visible of course) as being a target
• Anyone who tweets at someone who is a target and triggers the Bayesian filter gets auto-suspended and has to submit a form explaining why they should be unblocked