Transformation of metrics used in clusterization methods for building the phylogenetic language trees

As large typological databases appeared a few years ago, the problem of data mining (as clusterization of languages) arose. Usually phylogenetic algorithms based on Hamming-distance are used for these purposes. But it was found out in cluster analysis that some other metrics give better results. In the paper two new metrics are proposed and it is shown on a great number of linguistic examples that phylogenetic algorithms based on these metrics give better results.