Topic Modeling in Sociological and Political Studies

On March 4, was held the regular seminar of the "Machine Learning and Social Computing" research group. Theme of the seminar: "Topic Modeling in Sociological and Political Studies". The event was attended not only by the members of the research group, but also by graduate students and teachers from different cohorts and programmes.

It is expected that the next part of the seminar will be held on this Saturday, March 11. However, after the first day, participants already received a solid theoretical knowledge and a common understanding of what is topic modeling and how to use it in research. The head of the research group Alexander Sirotkin told about the mathematical component of topic modeling, the properties of the Dirichlet distribution and the iterative process of constructing the model.

Further, Denis Bulygin conducted a practical part of the seminar. Together with Stanislav Pozdniakov, Vadim Voskresensky and Victor Karepin, they prepared a topic model of the UN news for 15 years with the keyword: "peacekeeper".

Denis showed step by step all the stages of topic modeling, from preprocessing of the data to decoding topics and analyzing of the results.

At the end, the seminar participants discussed in which projects they apply or plan to apply topic modeling. The second part of the seminar will focus on the integration of topic modeling into quantitative research and its application for the analysis of online communities.