Study reveals hidden fear and anger in global social media coverage of Ukraine war
Researchers analyzed 1.5 million multilingual tweets to map emotional patterns about the Russia-Ukraine conflict, finding fear dominates discourse in conflict zones while Western countries show polarized sentiment. The findings offer policymakers and media organizations tools to detect manipulation, track crisis impact, and understand how national narratives mask outlet-level bias.
Originaltitel: Analyzing Emotional Discourse in Multilingual Social Media: A Case Study of the Russo-Ukrainian Conflict
<p>In this paper we propose a comprehensive analytical pipeline that efficiently processes 1.5 million multilingual tweets about the Ukraine-Russia conflict, relying on multilingual NLP techniques and strategic sampling to produce interpretable sentiment and emotion visualizations. By incorporating temporal and geospatial metadata as multimodal information, our approach reveals complex emotional patterns: filtering predominantly neutral content highlights fear (often over 60%) and anger (10–25%) as dominant emotions, with temporal analysis showing negative sentiment surges during crises. Geographically, conflict zones are fear-dominant, while Western democracies and state-influenced systems exhibit mixedemotional profiles; visualizations further demonstrate how national data can obscure media outlet polarization. Crucially, we enhance explainability by analyzing confusion matrix patterns to identify emotionally ambiguous or surprising predictions, such as fear being frequently assigned to neutral or positive content indicating pervasive underlying emotional currents. The outcomes of our work offer advanced, multimodal, and interpretable tools for analyzing large-scale discursive phenomena.</p>