I've just noticed the GNU diction package, which applies a number of "figure of merit" measures to text. Although the figures it generates are of debatable usefulness, it might be an interesting experiment to run it on a Wikipedia dump
Some questions: * Do the various readability scores rise as articles are revised? * Would it be worth creating special pages to find the best/worst articles by reading score, so they can be copyedited, or checked to see if they are suitable for being featured articles?
-- Neil