Parser Showdown at the Wall Street Corral: An Empirical Investigation of Error Types in Parser Output