Extracting Location-stamped Events from Textual Data for Persistent Situational Awareness

ABSTRACT: Air Force intelligence analysts need the ability to more rapidly monitor, visualize and analyze event information in large volumes of unstructured textual data. While research has been done in this area before, there is still a need for much higher accuracy extraction and geocoding of events. We propose to develop technology for event location stamping by combining key components in a way that has not been done before. Specifically we propose to improve the state of the art in location detection, location resolution, and trajectory resolution. BENEFIT: Current geotagging systems frequently fail to assign location stamps to events or assign incorrect location stamps to events. In both cases, visualization systems that present information as maps cannot perform properly. This project will result in the development of a fully automated event extraction and location-stamping system that is able to reason about time and assign correct location stamps to all extracted events. The solution will have broad commercial potential wherever knowledge workers perform geospatial analysis of unstructured text. In addition to software applications for intelligence gathering and analysis in defense, homeland security, and law enforcement, commercial potential exists in knowledge-based businesses such as professional services, engineering, medicine, and law. First commercial applications should be targeted where knowledge workers are faced with large quantities of unstructured text necessitating a geospatial analysis, for example disease discovery and tracking.