Text Classification by Genres Based on Rhythmic Characteristics

This article considers the rhythm of texts of various genres: fiction novels, advertising, scientific articles, reviews, tweets and political articles. The authors identify such lexical and grammatical means in the texts as anaphora, epiphora, diacope, aposiopesis, etc., which are markers of the rhythm of a text. On their basis, statistical characteristics are calculated that describe quantitatively and structurally these rhythmic means. The resulting text model is visualized for statistical analysis using boxplots and heatmaps, which shows differences in the rhythm of various genres. The boxplots shows that almost all genres differ from each other in terms of the overall density of rhythmic characteristics. The heatmaps shows the different rhythm structure of the genres. Further, rhythmic characteristics were successfully used to classify texts by six genres. The high quality of the classification shows that rhythmic characteristics are a good marker for most genres, especially for fiction. The experiments are carried out using the ProseRhythmDetector software for Russian and English. Text corpora contain 300 texts for each language.