NLP-Based Microblogging Post Filtering for Disaster-Related Content

During times of tragedy, when the usual channels for emergency communication are overwhelmed, people are forced to utilize any means at their disposal in order to reach out and seek assistance from others. In case of a disaster, it's crucial to precisely gather and share information to reduce the impact. Twitter has evolved as a medium, to share information during disasters and the widespread usage of cellphones enables people to share information about crises in real time. As a result, lots of organizations such as rescue groups and media organizations are more intrigued, in monitoring Twitter. However, determining whether a person's statements truly indicate the occurrence of a disaster is not always straightforward. The study explores how advanced natural language processing techniques, Transformer models are used to detect and classify emergency messages, in tweets related to disasters. Our method achieves a 96% accuracy showcasing its reliability and effectiveness, in real world scenarios.