Presentation @ Uni Trento: Quantifying Perceptions of Violence on YouTube
This study presents an innovative approach to quantifying perceptions of violence by leveraging geolocated YouTube comments. By developing a granular, spatially and temporally resolved dataset, we aim to measure perceived violence in near real-time and examine its correlation with public events. Our methodology includes identifying relevant videos using the YouTube API, analysing their comments for violent content through a dictionary-based approach, and aggregating this data across geographic grids. These measures are fine-tuned to reduce systematic biases, enabling real-time sensing and forecasting of societal tensions. This work not only enhances the understanding of how violence is perceived across communities but also offers tools to mitigate misinformation and protect vulnerable populations.