Code and data for "Neural topic modeling reveals German television's climate change coverage"

This is the code repository for the manuscript "Neural topic modeling reveals German television’s climate change coverage" by Schirmag, Wedemeyer, Stechemesser and Wenz published in Nature's Communications Earth and Environment. The study analyzes climate change coverage in Germany's largest television news program, Tagesschau. The media, and television in particular, play a crucial role in informing the public discourse and therefore shape the public perceptions of climate change. To analyze the climate change coverage in Tagesschau, we developed a custom news story identification algorithm that extracts the individual news items from the Tagesschau subtitles by combining several NLP methods. We classified the topics of all news items by combining a dictionary-based approach with neural topic modeling to study the themes and prevalence of climate change coverage. All related code is available here. Additionally, we provide a Jupyter Notebook to replicate our analysis results, as well as the necessary data. To get started there are some helpful hints in the README file. For more details please check out the article, as well as the Supplementary Information here:

Schirmag, T., Wedemeyer, J. H., Stechemesser, A. & Wenz, L. Neural topic modeling reveals German television’s climate change coverage. Communications Earth & Environment 6, (2025). https://doi.org/10.1038/s43247-025-02402-1