Social pressure analysis of local events using social media data

The lack of access to clear and actionable information and analysis to law enforcement agencies during the “Unite The Right” rally in Charlottesville and the torch-lit march the night before (August 11–12, 2017) was found to be a major handicap in the handling of the situation. To address these issues, this study analyzes online activity associated with events such as this on an ongoing basis and can be provided to local police departments so that they can more effectively monitor information on emerging events and respond appropriately. In this work, we introduce the concept of social pressure to assist human users to identify and track trends that may lead to potentially violent events. We combine existing methods of analyzing social media data for event detection with monitoring of the social pressure that may lead to such events. Our algorithm detects words and phrases appearing on social media that may be of interest due to their pertinence to real-world events or movements. After identifying words and phrases that may correspond to news or events, the social pressure is interpreted from the Latent Dirichlet Allocation topic weights and sentiment scores of the tweets over time. The resulting algorithm is able to consistently detect keywords related to events before they occur, and provide valuable insight into the nature of the events.