AI-Driven Automated Crisis Distress Classification System using Natural Language Processing

The social media platforms have been of great essence as sources of real time information in case of disaster and emergencies. People tend to provide immediate updates on the necessity of rescue, medical emergency, infrastructure damage, and resource shortages using social networks like Twitter. Despite the usefulness of the situational information in these messages, the emergency response teams find it challenging to monitor the large amount of unorganized social media data manually. The proposed research is a crisis distress classification system based on artificial intelligence, which automatically analyzes the messages in social media to determine disaster related information and prioritize distress signals. The suggested framework employs the BERTweet transformer model that is specifically created to interpret informal language that is widespread in social media posts. The pipeline being used in the system processes tweets in the Kaggle Disaster Tweets dataset by using a modular pipeline that includes text preprocessing, generation of contextual embedding, crisis classification, urgency prediction, and location extraction via Named Entity Recognition. The crisis messages are divided into the classes of disaster-related and another classifier forecasts the urgency levels to prioritize the response activities. The organized outcomes are combined with the UiPath Robotic Process Automation workflow, which automatically works with the output and sends email alerts on emergency monitoring. The offered solution emphasizes the possibility of integrating the transformer-based Natural Language Processing with the automation technologies to help to provide intelligent disaster management systems and real-time monitoring of the crisis.