TEMA
TEMA (Trusted Extremely Precise Mapping and Prediction for Emergency Management) is a four-year long project funded by the European Union flagship research and innovation program Horizon Europe. The project is carried out by a consortium of 19 partners across Europe and aims to improve Natural Disaster Management (NDM) through extreme data analytics. When a natural disaster like a flood or wildfire occurs, data analysis needs to be both extremely precise and rapid. TEMA aims to address this gap by delivering AI-enabled NDM platform that allows for the precise observation and modeling of real events through multiple data sources. In this context, geo-referenced social media data (geo-social media) has become a vital data source for disaster managers as it can provide real-time crowd-sourced insights that might not be covered by official reports. The Geosocial AI group at IT:U aims to improve the utility, accuracy and speed of algorithms for geo-social media data in a disaster context. Specifically, we have developed a multimodal method to extract what people post where and how they feel. We have also developed a model for aspect-based emotion analysis which can extract emotion regarding specific aspects of short texts. Additionally, we explored the capability of social media data for wildfire detection and improved relevance classification through a multimodal framework that integrates space, time and text. Our algorithms are being integrated in the TEMA technical platform tested in eight real-world trials together with disaster managers, covering floods and wildfires in Italy, Germany, Finland and Greece.