Using computational methods, digitized collections and archives can today be scrutinized in their entirety. By distant reading and topic modeling one particular collection – 4500 digitized Swedish Governmental Official Reports (SOU) from 1922 to 1991 – this article gives a new archival perspective of the history of Swedish film politics and policy-making. We examine different probabilistic topics related to film (and media) that the algorithm within the topic modeling software Mallet extracted from the immense text corpora of all these Official Reports. Topic modeling is a computational method to study themes in texts by accentuating words that tend to co-occur and together create different topics. Basically, it is a research tool for the discovery of hidden semantic structures, exploring a collection through the underlying topics that run through it. Hence, our article captures a number of film discourses and trends within the SOU material. In conclusion, we argue that topic modeling should be recognized as a method and research aid for gathering an overview of a major material; as a way to pose new and unforeseen research questions; and as a kind of computational support that makes it possible to apprehend major patterns more or less impossible to detect through a traditional archival investigation.