MEMORY OF THE GREAT PATRIOTIC WAR IN COMMENTS TO REGIONAL MEDIA: AN EXPERIENCE OF COMBINING MACHINE LEARNING AND DISCOURSE ANALYSIS; ПАМЯТЬ О ВЕЛИКОЙ ОТЕЧЕСТВЕННОЙ ВОЙНЕ В КОММЕНТАРИЯХ К РЕГИОНАЛЬНЫМ ИНТЕРНЕТ-СМИ: ОПЫТ КОМБИНАЦИИ МАШИННОГО ОБУЧЕНИЯ И КРИТИЧЕСКОГО ДИСКУРС-АНАЛИЗА

This article is devoted to the study of the comments to regional media of Kaliningrad, Tula, and Tomsk regions about commemorations of the Great Patriotic War of 1941—1945 in light of a combination of Machine Learning (ML) methods and critical discourse analysis (CDA) methodology. The sources used were the comments of regional media users in the period from 2015 to 2021, dedicated to commemorations of June 22 and May 9, the “Immortal Regiment” action, as well as family memory of the war events. The analysis of a sample set of the comments using ML methods showed that in three selected Russian regions, regardless of anniversaries, the number of positive comments is decreasing, and the number of negative comments is growing. The Tomsk region showed the highest percentage of critical comments regarding all studied commemorations. The study made it possible to reveal the internal heterogeneity of this trend, showing the relationship between the use of images of the war and the assessment of the current socio-political situation. The analysis showed that the use of memory of the events of the Great Patriotic War as a variant of civil religion in modern Russia faces the sacralization of individual heroes and images of family memory of the war, which are increasingly contrasted with the official narratives about the Great Patriotic War. At the same time, the study showed that criticism of commemorative events is more related to current regional socio-economic problems and is extremely emotional and situational. This conclusion made it possible to reveal in more detail and substantiate the thesis about the presence of tendencies of polarization and fragmentation of mass historical consciousness in the regional Internet space. Based on the examples studied, the principles for combining ML methods and CDA were formulated. © 2025 Elsevier B.V., All rights reserved.