zlacker

>>pg+(OP)
3 words: Bayesian Comment Filter. Just does the opposite of what the spam filter does. Use the corpus of great comments from the past to find great comments of the present.

I'm only half joking. Fundamentally, the thread is about a filtering system.

>>b_emer+A4
The problem is that word features are not really that good predictors of quality.

I have done some research on this (unpublished), and I got a really good performance on predicting hacker news votes by just counting how many new words (not stopwords, not very-high-frequency words) a comment was adding to a thread. Just using a few variations on this theme predicted better than word counts or bigram features.

Fundamentally, though, I disagree with machine learning- based approaches as they can only _reinforce_ present behavior, and we'd like to shape voting behavior.