Modeling Media History: On topic models of Swedish media politics 1945–1989

In an explorative manner, this article uses a data-driven digital history set-up to focus on media political issues in Sweden during the second half of the twentieth century. By distant reading and topic modeling a dataset of 3100 Swedish Government Official Reports between 1945 and 1989—a corpus of some 87 million tokens—the article gives a new perspective of how the Swedish state examined and discussed media in general and media politics in particular. Topic modeling is a computational method to study latent themes or discourses in a dataset by accentuating words that tend to co-occur and together create different topics. Via a computational interrogation of the dataset in a Jupyter Lab environment a number of media topics can be detected. They include the most common words for each media topic, but also reveal temporal periodizations when media political issues were foremost discussed as well as other societal topics that media was related to.