Automatic search of indicators of text authorship

This paper is the work in the field of text authorship analysis-stylometry. The main assumption underlying stylometric studies is that all ripe authors of texts have an unconscious aspect to their style. Stylometrics use enormous variety of text features for determining authorship of texts. But there are no definite rules from which indicators to choose and finding ones for a specific set of texts seems to be a black art. In our work we have tried to realize the automatic feature extraction algorithm based on the genetic search. The special code allowing carrying out searching of indicators of text authorship and the function for determining quality of obtained variants have been developed. The details, results and conclusions of the experiment are presented.