An LLM driven dataset on the spatiotemporal distributions of street and neighborhood crime in China

Abstract Crime is a significant social, economic, and legal issue. This research presents an open-access spatiotemporal repository of street and neighborhood crime data, comprising approximately one million records of crimes in China, with specific geographic coordinates (latitude and longitude) and timestamps for each incident. The dataset is based on publicly available law court judgment documents. Artificial intelligence (AI) technologies are employed to extract crime events at the neighborhood or even building level from vast amounts of unstructured judicial text. This dataset enables more precise spatial analysis of crime incidents, offering valuable insights across interdisciplinary fields such as economics, sociology, and geography. It contributes significantly to the achievement of the United Nations Sustainable Development Goals (SDGs), particularly in fostering sustainable cities and communities, and plays a crucial role in advancing efforts to reduce all forms of violence and related mortality rates.