MINGLER: Multi-modal Information GLEaner for Contact Report Filtering

The basic concept of a contact report has not changed significantly since the advent of modern warfare. Communicating the location, type, heading, position, and other information of possible enemy entities is a fundamental part of intelligence gathering in both peacetime and war. What has changed drastically is the quality and quantity of reports that are generated and provided to intelligence analysts and interpreters. While many of these reports contain relevant information, most reports received by analysts are either irrelevant, ill-formed, or misleading. As a result, analysts are now inundated with too much meaningless information, and their throughput suffers as a result. CoVar’s prior work (HERMIT) enables us to extract temporal/geo-spatial distributions from arbitrary data sets and project data from arbitrary modalities into a unified feature space for entity tracking. With slight modifications, these key technologies naturally enable large-scale data filtering and thus can help build human-machine teaming AI/ML software for contact report filtering.  MINGLER (HERMIT for Contact Reports) will automatically filter bad contact reports from arbitrary sensors or sources and empower analysts to turn individual scattered contact reports into a coherent understanding of the world.