Newsreels, short news films shown in the cinemas in the 20th century, had an impact on how the audience saw the world, and what they thought was worth knowing of it. Even when the audience was critical about the newsreel contents, they could see from the newsreels what was supposed to be the mainstream view to the contemporary world (Chambers, Jönsson & Vande Winkel 2018; Imesch, Schade & Sieber 2016). Studying newsreels is crucial, because understanding the worldviews that newsreels conveyed in a long temporal perspective would allow to explain better how audiovisual influencing and propaganda works, and how that has changed over time. Due to the large amount of data, systematic exploration of newsreels covering several decades requires using computational quantitative methods alongside critical qualitative analysis.
This talk presents an ongoing study that explores how to best study patterns and dynamics of depicted worldviews in a large digitized corpus of historical newsreels, and to what extent that is at all possible. We explore what are meaningful and traceable units of analysis, and how to take into consideration the kaleidoscopic nature of audiovisual news in a computational study. The research is based on the multidisciplinary approach inherent to Cultural Analytics. We take inspiration from preceding computational research on newsreels (Carrive et al 2021; Heftberger 2018; Althaus et al 2018), cultural history of knowledge (Burke 2015), the Distant Viewing framework (Arnold & Tilton 2019), and approaches proposed by Lev Manovich (2020). We use the methods of machine learning of visual images and multidimensional embedding spaces, Natural Language Processing, social network analysis, temporal dynamics, and geographical mapping in our study.
The data used in the study consists of digitized newsreels produced by the Central Documentary Film Studios in Moscow in 1944-1993 from the collections of Net-Film company.