Newsreels, i.e. short news films shown in cinemas right before the main feature, served as illustrators of the news provided by newspapers and radio to wider audiences in the 20th century. Exploring large collections of newsreels opens unprecedented possibilities to systematically analyze long-term cultural phenomenae, such as representations of ‘facts’ and their changes. Employing a cultural-data analytics approach in multidisciplinary collaboration of Mila Oiva, Vejune Zemaityte, Ksenia Mukhina, Andres Karjus, Daniel Chavéz Heras, Mikhail Tamm, Tasweer Ahmad, Tillmann Ohm, Mar Canet, and Maximilian Schich, provides new insights to the data and therefore opens possibilities for challenging the existing perceptions of informational audio-visual culture. Despite the recent developments of machine learning and ‘distant viewing’ of audiovisual materials, the conceptualization of computational analysis of audiovisual heritage data is still underdeveloped. In order to respond to this, the CUDAN Newsreel collaboration develops a new framework for exploring a large cultural heritage collection of Soviet newsreels to characterize the continuities and changes in the official Soviet representations of world events. The project develops and exemplifies the proposed methodological framework via the Soviet newsreel journal Daily News (Новости дня) published weekly in 1944-1992, and digitized by Net-Film company (https://www.net-film.ru/). Driven by and integrated with strong expertise in cultural history, the framework entails a data science workflow pipeline, including audiovisual machine learning methods, as well as is complimented by insight from film studies and creative industries disciplines. The research questions of the project concern the temporal patterns that can be identified through the exploration of places or organizations depicted in newsreels, the composition of film production labour networks, and the aesthetic features of footage.
Conference website: https://necs.org/