Seminar in Computational Linguistics

  • Date: –14:30
  • Location:
  • Lecturer: Raazesh Sainudiin
  • Contact person: Artur Kulmizev
  • Föreläsning

The Polarised State of the Swedish Political Twitterverse: Lessons from Ideological Forests of Hate in the 2016 US Presidential Election


I will spend the first 30-40 minutes on Characterizing the Twitter Networks of Prominent Politicians and SPLC-defined Hate Groups in the 2016 US Presidential Election, Social Network Analysis and Mining, 9:34, 2019.
Then I will overview a preliminary analysis of over 91 million tweets collected during an 8-month period around the Swedish general election in 2018. This analysis showed strong evidence of highly polarised communities. These communities were found to differ politically and in their nearly mutually exclusive use of hashtags, links to URLs and preference for news sources. This preliminary study of the Swedish Twitterverse in conjunction with statistical tests is indicative of highly polarised echo-chambers across the left and right political spectra. 

This mathematical research has mainly focused on NLP-free network signals in the social media communications data towards testable models of empirical ideological processes (eg. echo chambers) in social media communications. My main objective with the seminar is to inspire NLP researchers to exploit the unique dataset on the Swedish political twitterverse and brainstorm pathways for future collaborations.
Affiliations: Associate Professor of Mathematics with Specialisation in Data Science, Department of Mathematics, Uppsala University, and Director, Technical Strategy and Research, Combient Mix AB, Stockholm
Acknowledgements: This work was done jointly with Agnes Davíðsdóttir, Magdalena Fischerström, Claes Fälth, Johannes Graner, Andreas Lindgren, Amela Mehic and Albert Nilsson and Tilo Wiklund as part of the Uppsala Summer Math Camp, Department of Mathematics, Uppsala University, and sponsored by AWS, databricks and Combient Competence Centre for Data Engineering Sciences.