Blue Witness

Incident Reports Grouped by Level of Police Force

The Twitter scraper on the Blue Witness project is responsible for gathering tweets with content-specific criteria, including (but not limited to) keywords such as “police” and “misconduct” and their synonyms. It does this multiple times a day and checks the incoming data against data in our database to prevent duplicates from being entered into the database. It then passes them to the BERT model to be classified and awaits administrator approval. The Twitter bot is a tool that reaches out to Twitter users to confirm or supply the missing information. It is also triggered by the Administration Team, and new data gathered awaits approval before being marked as an ‘approved’ incident report. This enhances the Blue Witness administrator’s ability to quickly update the incident map as well as the data available for download for researchers, journalists, students, and activists. If you’d like to know more, feel free to reach out to HRF on our contact page.

Project Details

The Blue Witness project is a movement to give the public access to reports of police use-of-force incidents. By crowdsourcing incident reports from Twitter, we create a platform where people can contribute to a greater cause. Our database allows our users to export report data for their use. We strive to provide a single source of truth on the topic of police use-of-force, that the public can use to inform themselves in the current state of the issue.


What We Did


Analyzing data isn’t easy, let me give you a fresh pair of eyes!