A probabilistic evaluation of similarities among very dissimilar languages