Despite its rather modest number of users, Twitter has gained a certain status as platform for on-time political communication and recent studies have brought evidence, that signicant events tend to produce peaks in Twitter activity. This article focuses on the interaction between Twit-ter and traditional mass media channels (newspapers, magazines, TV) by analysing topics central in Twitter communication on politics. The data-driven analysis presented in the article analyses the structure and dynamics of topics discussed on Twitter during elections. The data used in this article are from two political events, the German Federal Elections 2017 and European elections 2019. Both datasets contain more than one million tweets Empirically, the article seeks to evidence the usefulness of Twitter data when it comes to gain understanding of actual political debates in a society. Methodically, the article seeks to present how data mining, modelling and visualisation tools can be exploited in order to extract scientically relevant, reliable and signicant, and analytically interesting results from a very large dataset to be used to answer questions being relevant for political and social sciences and media studies as well.