How to measure political polarization in text-as-data? A scoping review of computational social science approaches

The rise of political polarization within western societies has been portrayed by events such as the United States Capitol riot or the United Kingdom's exit from the European Union. In this context, we argue that computational social science (CSS) methods offer a scalable and language-independent fashion to measure political polarization, enabling the processing of big data. In this vein, this article offers the first scoping review of the application of CSS methods to analyzing political polarization through text as data. We propose a categorization framework and reflect on the advantages and disadvantages of the different models used in the literature. Additionally, we underline the importance of filling research gaps, such as considering the temporal characteristic of political polarization, using a mathematical approach to analyze the use cases, and avoiding location and platform bias. We also provide recommendations for future research.