«Особые приметы» в речи художественных персонажей: количественный анализ диалогов в «Войне и мире» Л.Н. Толстого
This paper presents a quantitative study of spoken dialogue in Leo Tolstoy’s War and Peace. Tolstoy was known to put a lot of emphasis on the language in which fictional characters express themselves, and conscious modification of their speech is acknowledged by critics as part of his literary technique. Our goal was to try and find some formal markers that would help us distinguish the characters, measure some sort of speech-based similarity between them, and cluster them into meaningful groups. At the first stage we applied some well- established approaches of stylometry (computational stylistics) that were originally developed for real-world authorship attribution and rely mainly on word and n-gram frequencies. Then we tried our own alternative method based on more formal and structure-oriented features independent of actual word choice. Both approaches produced meaningful and interpretable results, which indicate overall applicability of quantitative methods to literary studies in general and to the analysis of specific characters in particular. At the same time, the difference between the two sets of results helped us demonstrate that sometimes more formal and structure-oriented features could be more revealing and ‘noise-resistant’ than word and n-gram frequencies.