I'd be interested in a breakdown of the 10 or so most popular source sites on a (say) month-by-month basis. You'd expect to see a lot of articles referencing Google, techcrunch, HN and the like but I'm also surprised by the number of articles from the NY Times, WSJ and such.
Maybe determine "popularity" first by number of articles and then again by score.
* Objectively ranked categories based on word usage, not author-provided tags
* Legible graphs with fewer colors
* Analysis that weights comments by karma (or, better, a reasonable non-linear function on karma)
* Open source for reproducibility and better outside critique
Edit: In fact, I'd like to see it enough that I might build it. Any other ideas?