PUBLIC AGENDA AND PUBLIC OPINION IN THE RUSSIAN BLOGOSPHERE (2012-2013)

This two-year research was the main component of the Lab’s project supported by the Basic Research Program of NRU HSE in 2012 and 2013. It built upon the previous project aimed at development of methodologies for sociological analysis of the blogosphere. The goal of this research was to describe topical structure of the blogosphere exemplified by the Russian-language LiveJournal and the relationship of this structure to other parameters of blogs. The data for the project were automatically downloaded from LiveJournal with BlogMiner software developed in the Lab. During the two years, about one hundred relational databases were collected that contain together almost four million posts and approximately 20 times more comments, as well as data on their connection, dates and authors’ IDs. Another Lab’s software, TopicMiner was used in the project to automatically retrieve topic structure from large collections of texts with the algorithm known as latent Dirichlet allocation with Gibbs sampling. Once retrieved, topic structure was related to other parameters.
In particular, it was found out that overall topical structure does not change in time, with overall weight of public affairs topics being approximately equal to that of private and recreational topics. However, inside the public affairs group there exists the most volatile cluster of event-centered topics which was most visible in comparison of periods before and during the national elections 2011-2012. It has been also established that the topical structure of popular bloggers’ posts does not significantly differ from that of ordinary bloggers. At the same time ordinary bloggers are characterized with much lower activity and with the presence of noise created by spammer accounts. While studying other parameters, we found out that the number of received comments only weakly correlates with the number of generated posts, which gives a possibility to construct an index of bloggers’ efficiency by calculating mean number of comments per blogger’s post. Though blogs contain multiple topics, in several blogs certain topic groups prevail that make it possible to create thematic profiles of bloggers, and to cluster bloggers according to these profiles. The research also revealed the variation range of bloggers’ activity according to the day of the week and time of day (on weekends the activity decreases by a quarter), which helped calculate adjustment coefficients for the correct identification of activity peaks.