Web content mining analysis of e-scooter crash causes and implications in Germany

In Germany, police reports published via press are neither uniformly written nor accessible to the public. There is a lack of comprehensive and factual data-based analyses of e-scooter crashes and their causes. We collected 1936 crash-related reports over two years via the German press portal based on a systematic web content mining process. Sentiment analysis results revealed that the police reports' coverage is predominantly factual and neutral and, therefore, useful for keyword-based analyses. After identifying the 46 most relevant keywords in the reports, we generated an adjacency matrix to investigate the keywords' dependencies, visualized the network and dependencies of the most relevant keywords, and categorized them into four thematic clusters using the Louvain algorithm. Our results and findings reveal that driving under drug influence, especially alcohol, is one serious problem. Riding e-scooter in pairs and on forbidden terrain or in the wrong direction are also common causes of crashes. Consequences for e-scooter riders are severe injuries, driving license revocation, fines, criminal charges, and incurring for property damage. Further, wearing protective gear and helmets is of low acceptance among the e-scooter ridership. Based on our results and findings, we recommend e-scooter bans during the night times for some locations, obligatory driving tests before first e-scooter use, and helmet wearing.