[ SmokeDetector ] URL in title, bad keyword in body, bad keyword in title, link at end of body, link following arrow in body, pattern-matching website in body, pattern-matching website in title: trexmusclesite.com/optimum-garcinia-plus-reviews/ by fenting on drupal.stackexchange.com…
> Pattern-matching website in body, pattern-matching website in title ---------- Title - Position 14-28: .trymaxface.co Body - Position 14-28: .trymaxface.co
Still, that file's not as long as I thought it might be, and the scripts don't take as long to process the stuff as I thought. So far, this is going well.
Take all the stuff we've got in metasmoke in one massive JSON set, pick it apart, do some text learning on it.
@Undo individual words, but not individually :) Basically, work out the probability of a word occurring in spam and in ham. Then for every word of a post I check, find the probability record and... math. Somehow.
Can someone run me through how bodyfetcher works? How do we make sure we're not fetching bodies two hours after the post, when it should have been fetched in the first few seconds?
@ArtOfCode On sites other than SO, questions (with answers) are fetched when they come up in the realtime websocket. So they are fetched with every edit.
On SO, we still look at the websocket, but the fetching process is to ask for the last N questions (by activity) once M accumulate in the queue. Here N > M because not all SO questions go through realtime.
@ArtOfCode Yes, the queue is there as well, so that on most sites a couple of posts are fetched at once. But this doesn't change the cause-effect relation of websocket - fetch.
Ok, I shouldn't have said "fetched when they come up". They are queued up when they come up in realtime. And fetched when queue size reaches a threshold.
Each site has its own queue, and threshold is also site-specific.
> Would not be caught for title, body, and username.
> Bad keyword in body, bad keyword in title, bad keyword in username ---------- Title - Position 1-7: hitler Body - Position 1-7: hitler Username - Position 1-7: hitler
@SmokeDetector: It seems this is the fourth time that it has seemed like it was my first time sending feedback. Make sure you're persisting the database where you keep this information.
> Blacklisted username, offensive body detected, offensive title detected ---------- Title - Offensive keyword: niggers Post - Offensive keyword: niggers Username - Position 1-7: nigger
