Social Media Analytics for Disaster Management using BERT Model

Disasters, both natural and human-induced, continue to affect the global population in multiple ways. In the past decade, social media has evolved and has seen a rapid, exponential increase in the number of users and their data. The manner in which communities anticipate, react to, and recover from natural and man-made disasters have been entirely shifted by rapid information transfer, public sentiment analysis, and real-time communication. There is now a high demand for social media based disaster management systems. This research project introduces a sentiment analysis framework tailored to Twitter data, incorporating BERT as a core component. Additionally, we employ a range of NLP methodologies for evaluation and comparative analysis. The results indicate that the employment of BERT yields a noteworthy enhancement in accuracy, surpassing alternative models by 5% to 7%. In summary, our research accentuates the effectiveness of employing BERT-based sentiment analysis to optimize disaster management strategies.