Listening for Ideology: Automatic Analysis of Character Speech in Historical Nazi Propaganda Films

Abstract

While the visual dimension of film has been widely explored in digital humanities through methods such as "distant viewing", the audio layer has received less attention despite its crucial role in meaning-making. We address this gap with a four-step pipeline combining speaker diarization, audio gender classification, automatic speech recognition (ASR), and LLM-based psycholinguistic analysis to infer character traits from film dialogues. Applying this method to a set of Nazi propaganda films, we find that despite challenges in speaker diarization due to noisy historical film audio, modern ASR and GPT-based analyses produce character profiles consistent with existing filmic research. Our proposed pipeline advances distant reading of film dialogue, complementing visual analyses and enabling scalable study of ideology in historical cinema. A case study of female characters in NS films identifies three recurring types, centered on the ideological figure of the mother in National Socialism.

Resources

Details