I think this algorithm is giving too much weight to the html tags. Html tags wouldn't constitute actual thought, would they? Even though they are content.