Forensic Intelligence Derived from Crime Scene Evidence Using Text Embeddings

Forensic Intelligence and Crime Linkage are fields dedicated to studying patterns in criminal behavior to provide actionable insights for investigations. Despite their potential, both fields have seen limited advancements in leveraging recent developments in Artificial Intelligence. This study introduces a novel model that utilizes Natural Language Processing techniques to analyze criminal behavior patterns based on crime scene evidence. Our methodology involves transforming detailed crime scene narratives into a vectorized format, creating a multidimensional space that encapsulates the modus operandi of individual cases. When applied to a real-world dataset, preliminary results demonstrate the effectiveness of this approach in grouping similar crimes, including known serial offenses, and distinguishing unrelated cases based on location and forensic evidence analysis. This research aims to benefit both the academic and law enforcement communities, providing a tool that uncovers insights into criminal behavior patterns and potential case linkages.