A Data Generative Approach to Criminal Justice Records Linkage Problems

Law enforcement is a multi-stage process, but law enforcement agencies, unfortunately, adopt heterogeneous reporting procedures when collecting different datasets. To reflect this multi-stage law enforcement process, we are working on a data-generative approach catered to large-scale criminal­ justice-related administrative records to harmonize and link the data from different stages. Prior works mainly focused on linking administrative records with static information like name, date of birth, and home address. However, law enforcement data often contains temporal information like the time of 911-reporting, police dispatching, etc., which needs further research efforts to model and link them accurately and reliably. The stitched datasets will provide a more systematic characterization and insightful understanding of law enforcement that is hard to imagine otherwise.