I'd be more interested in the complement -- filters tuned to pick up smart or interesting writing. I'm not convinced that it's necessarily an identical problem.

I'd be more interested in the complement -- filters tuned to pick up smart or interesting writing. I'm not convinced that it's necessarily an identical problem. It would be neat to run a battery of standard semantic analysis tools against the text of web pages ranked highly on HN, compared with pages not ranked highly.