Ranking mechanisms for word similarity