Cultural representations typically contain illuminating biases. For example, geographical locations are unequally portrayed in media, creating a distorted representation of the world. Identifying and measuring such biases is crucial to understanding both the data and the socio-cultural processes behind them. Here we measured geographical biases in the representation of cities in a large Soviet news-media corpus, highlighting relative emphasis. Combining urban science, cultural geography and digital humanities, we first obtained robust quantitative estimates of representational biases and then qualitatively interpreted the classifications, feeding these interpretations back in an iterative feedback loop. Applied to a corpus of Soviet newsreels, 'Novosti Dnya' (News of the Day, short news films shown just before feature films), we find that city representation grows superlinearly with city size and is further biased by city specialization and geographical location. For example, cities with major hydroelectricity and steelworks were overrepresented, while those in the industrial heartland and in non-European socialist countries were underrepresented.