Pelevin vs Sorokin: an Attempt of Stylometric Comparison
Andrei Viacheslavovich Zenkov, Miroslav Andreevich Zenkov, Nikolai Andreevich Zenkov · Филология научные исследования · 2024
Our study is related to quantitative linguistics and focuses on the application of a new method for analyzing the author's style in literary texts. The method uses computer analysis of numerical data found in texts, including both cardinal and ordinal numerals, expressed both in numbers and verbally. Author used the program which automatically removed phraseological units and fixed combinations accidentally containing numerals. Before analysis, the text must be manually cleaned of numbers that do not contribute to the author's artistic vision, such as page numbers or chapter numbers. The analysis revealed that the use of numerals by an author in his/her texts is unique and individual, forming a characteristic feature that distinguishes texts by different authors. For the first time, a formal quantitative stylometric analysis is performed of the literary works by Victor Pelevin and Vladimir Sorokin – authors whose literary styles share many similarities when viewed through the lens of a traditional descriptive philological approach. To validate this methodology, we have also included the texts of four "impostor" authors in our analysis. It has been found that Pelevin's and Sorokin's texts differ significantly in their use of numerals. The data on occurrences of numerals in the texts were subjected to hierarchical clustering, which accurately divided the texts into groups based on their authorship. Since the clusterization results can be influenced by the choice of both metrics and clustering method, we tried various reasonable combinations of them to ensure the reliability of our results. Each time, the dendrogram would change only slightly. Thus, the clustering outcomes were found to be reliable. The proposed new method of quantitative linguistics, which is based on the analysis of numerals in literary texts, has the potential to successfully solve the stylometric problems, particularly related to the attribution of texts.