Machine Learning and Text Analysis for Political Texts (several projects)

Potential supervisors: 

Several projects available to students interested in working with a cross-disciplinary research group with both political scientists and computer scientists. This MSc project will be conducted under the umbrella of the project “Bias and methods of AI technology studying political behavior( and will suit students interested in research in natural language processing, machine learning, topic modelling and other topics related to analysis of political texts. Some suggested topics are listed below, please contact the supervisors for more details:

  • Classification of political text according to party and/or left-right positioning. Working with texts from the Swedish parliament, we are interested in characterising the language use of the different parties/coalitions. Can we train a classifier to distinguish the party affiliation of e.g. motions from the Riksdag? Some techniques that could be evaluated for this task is M-BERT (multilingual BERT), a deep learning framework for NLP. There is also an opportunity to work on other data, e.g. US senate speaches.
  • What are the most prominent topics for different time periods in the Swedish parlament? In our current research, we have investigated unsupervised learning models to compare word embeddings over different periods and parties. This project would attempt to complement those results by exploring topic models to see what the most dominant issues are during the given time periods. When did certain topics (immigration, the environment etc) become more dominant, or less so?

Other projects under the theme are also possible, please contact us if you are interested in other topics related to NLP applied to political data, not least as there is a general election in Sweden in 2022!

Contact: Moa Johansson moa.johansson(at) and Bastiaan Bruinsma (contact info will follow)


Date range: 
October, 2021 to October, 2024