thoughtcrumbs

Naive Bayes stupidity filter

Niki pointed me to Stupidfilter.org, where they're building a Bayesian classifier for stupid comments posted to blogs. It works like a rule-based spam filter, and the features they've proposed include capitalization (too much or too little), leetspeak, use of exclamation points, etc.

Makes me think my plan to classify politeness isn't so far fetched. Of course, they point out that "a sufficiently advanced [stupidity detector] would probably filter out the whole of human discourse."