Analysing Protest-Related Tweets: An Evaluation of Techniques by the Open Source Intelligence Team

The police’s Open Source Intelligence (OSINT) team is constantly looking for better ways to analyze large corpora with high precision. This paper presents the evaluation by the OSINT team of more advanced methods than currently used to extract information from tweets related to an upcoming large-scale demonstration. An initial interview revealed the current way of working. Next, more advanced machine learning techniques such as sentiment analysis and network analysis are used to create visualizations that would better suit OSINT’s needs. Finally, our proposed visualizations are evaluated by two OSINT analysts who state that the results are clear, actionable, and relevant, whereas completeness of information and privacy pose additional challenges.