Automatic AI-Supported Information Extraction in Natural Hazards Reconnaissance

Hazardous events and corresponding responses require the exploitation of information from reports in various modes (text, image, video, audio) coming from responders, affected population and officials. In addition to the underlying need for automated processing of such reports and information, we explore the potential of implementing AI services for information extraction and knowledge management. We demonstrate that we can transfer the practices and technology from the military reconnaissance to civilian applications, while integrating the technological advances from the AI-based object detection and Large Language Models, and providing a simple way to automate the processing of multimodal data to generate intelligence to support decision making.