The distances distributions between empirical triplet distributions are investigated. The accuracy estimation of these distributions is obtained depending on the length of the text. The method of author identification is examined on the broad class of literature texts. The stabilization length of triplet distributions is approximately equal to one half of the text without dependence on author and text length. The example of cluster method is given for E.I. Roerich philosophical texts.

empirical probability, minimal text length, author identification

Mathematical modelling in actual problems of science and technics