Authoritarian Communication at Scale: What Text-as-Data Reveals about Russia
Non-free regimes restrict public access to credible micro-data and can distort self-reported information. As a result, research inference often relies on signals embedded in text. News transcripts, official statements, and social-media posts become sources where narrative choices and language provide observable proxies for otherwise hidden processes. Recent reviews of research on Russia emphasise the limits of survey-based evidence in repressive settings and call for broader methodological toolkits beyond self-reports. In this presentation, I show how text-as-data methods can meet that need by analysing state television news transcripts alongside Twitter/X posts to study, respectively, the supply of mass-media communication and #NoWar activism in present-day Russia.
In the first part of the presentation, drawing on my PhD dissertation, I present results from an analysis of a large corpus of news from Russian state-controlled television to quantify tone and agenda over time. Using NLP for sentiment and a semi-supervised approach to country/topic labelling, validated against human labelling, I show that Vladimir Putin is presented disproportionately in positive coverage, that his visibility varies between domestic and foreign news, and that the salience and framing of Ukraine evolve alongside editorial changes. The focus is on scalable, transparent pipelines, including pre-processing, model selection, validation, and robustness, which make mass-media manipulations measurable at scale.
I then turn to Twitter/X to study digital activism around #нетвойне (#NoWar). Using a corpus of over 100,000 tweets from the start of Russia’s invasion of Ukraine in 2022 through the end of 2024, I trace when anti-war expression intensifies, how frames and emotions circulate across communities, and what distinguishes durable mobilisation from transient bursts.
Taken together, this presentation offers a template for text-as-data research in authoritarian settings. It combines validated measurement with design-based checks to show what the regime seeks to place on the public agenda and, drawing on social-media analysis, provides a case study of how publics mobilise around anti-war expression.
