Beyond Points on a Map: AI for Understanding Disaster Areas

What is it about?

When disasters like wildfires, floods, or storms happen, quickly knowing where the damage occurred is crucial for informed decision-making. Traditionally, mapping disaster impacts relies on pinpointing specific locations – like cities or buildings – on a map. But disasters rarely stop neatly at city limits, and reports often describe areas in complex and vague language using phrases like “between two towns” or “near the river.” Understanding the area impacted based on how it’s described in language is the task we call “Georelating.” This research introduces a new approach using artificial intelligence (AI), and more specifically, Large Language Models (LLMs), to tackle Georelating. We’ve developed a system that can read news reports, interpret the areas affected even if they aren’t clearly defined, and map them using a standardized grid system. Integrating geographical knowledge bases ensures grounding the interpretation on accurate spatial information. This allows for a more accurate and comprehensive picture of the impacted area, helping organizations make faster, more informed decisions. This work is a step toward building smarter disaster response systems that can quickly and reliably process global information.

Featured Image

Photo by Mike Newbry on Unsplash

Photo by Mike Newbry on Unsplash

Why is it important?

Current systems for pinpointing where disasters happen often miss crucial details. They rely on named regions and official boundaries, and ignore relational language that connects and specifies locations. Our Georelating system aims to address these limitations. This is important because it allows organizations to quickly track disasters around the world using news reports. By organizing this information in a standardized way, we can easily combine it with other vital data, like where supplies are needed or where infrastructure is located, helping them respond more effectively.