talideon.com

ERROR: #randomsig[27]# not found

October 8, 2004 at 4:33PM Bayesian comment moderation.

An idea’s been bouncing around in my head for a while now. Right now, I don’t have any comment moderation, and I wouldn’t like to have to build a moderation system, primarily not because of the difficulty--it’s not hard--but because I don’t like the idea of moderation from an ethical point of view.

Seeing as I’ve set up my site so that any pages with comments don’t allow spiders to follow links off them, link spam in comments shouldn’t be a problem. But there’s the possibility of spammers posting up crap anyway, and there’s also trolls to worry about.

I don’t want moderation, but if I was to build it, here’s how I’d do it: Bayesian moderation. After all, if it works for mail, why can’t it work for comments?

Like with email, when you post a comment it’ll run it against a filter. If it returns a negative, it’s posted up, otherwise it’s posted on a moderation queue for me to check. I’d be informed by email of the addition.

There’d also be moderation buttons on each comment where people could mark comments for moderation, be it as spam or trolls.

Ideally, the system would be able to do most moderation by itself, quietly disposing of spam and troll posts without me having to worry about doing it. If there’s any doubt about the spamminess or trollishness of a comment, then and only then would it be dumped on the queue for me to moderate.

It’d be nice to have two filter databases rather than the one, one for spam and the other for trolls. That way trollish comments wouldn’t dilute the spam database.

Technorati Search Technorati Search Irish Bloggers

Comments

No comments.

Post a comment

All form information is optional, but it’s a good idea to fill in your name and email address if you want me to take your comment seriously.

Spammers, don’t bother posting crap down here. The site is set up so that legitimate search engines (Google, for instance) won’t index pages with comments on them. Posting crud here only means you’re wasting my time and patience. Shoo!

Real names, please. Please include!
Won’t be displayed. Please include!
Displayed, if present.