The paper presents a way to assess the reaction to socially significant events in online discussions, based on statistically significant deviations (“outliers”) of numerical text indicators for some year’s observations. We consider outliers to be a marker of exceeding the natural variability of the diagnosed indicators, that is, an evidence authors’ psychological states correlated with the reaction to socially significant events affect textual activity. The TITANIS lexical and grammatical indicators of users’ reactions were evaluated in the thematic and background corpus of online discussions. Some ways of representation and interpretation are proposed for the patterns of statics and dynamics of outliers, thereby the severity and emotional orientation of users’ attitudes towards the topic of network discussion can be described.
Download the article (PDF) from eLibrary (in Russian, registration required): https://elibrary.ru/item.asp?id=82924296
Download the article (PDF) from the journal's webiste (in Russian): http://www.isa.ru/proceedings/images/documents/2025-75-3/68-79.pdf
Kuznetsova Yu. M., Stankevich M. A., Chudova N. V. Statistical outliers as a research object in the tasks of socio-psychological diagnostics // Proceeding of the Institute for Systems Analysis of the Russian Academy of Science, 2025, volume 75, № 3, pp. 68–79.