Frequency Portraits in Authorship Identification

摘要

Frequency methods of text content analysis are mainly used for decrypting cipher messages. In addition, these methods allow you to identify the features of both individual languages and topics. But not only for the language or topic, but also for each person is characterized by a personal set of features of speech behavior, which can be used to identify authorship. Methods of such analysis can be divided into two groups - expert and formal. Based on mathematical methods, you can create a model that differentiates one author of a text from another. This model is proposed to be called the Frequency Portrait. It includes single-and bigram frequency distributions, confidence interval, and probabilities. Algorithms and software have been developed to implement the frequency portrait model. Specific examples show the effectiveness of author identification. A method for forming a model of a set of single - and bigram frequencies for the author of text content and for a specific text is proposed. A system of terminology for such models is proposed. The concept of forming frequency portraits at various levels, from language to a specific text, is presented. The models of frequency portraits of the author of text content and frequency portraits of specific texts are presented. The possibility of identifying the authorship of text content on the basis of the presented concept is shown. The research was conducted within the framework of the Russian alphabet.