stylometry, syntax, genre, Czech, corpus

The objective of this chapter is to examine how dependency tags are distributed across various genres in the Czech language. Specifically, our focus is on identifying the similarities and differences in these distributions among different types of text. This study utilizes data from the SYN2020, a large balanced corpus of contemporary written Czech, comprising 100 million words. The findings indicate that the frequencies of dependency tags follow a power-law distribution, represented by the equation y = ax-b. When analyzing the values of the parameters a and b, a distinct pattern emerges that differentiates the genres. Genres within fiction, such as poetry, drama, novels, and short stories, typically exhibit lower values for a and b, whereas non-fiction genres, including administrative texts and professional literature, demonstrate higher values. Journalistic texts, like newspapers and leisure magazines, fall in between fiction and non-fiction literature in terms of these parameter values. Consequently, comparing these parameters appears to be an effective method for conducting stylometric research.