Simple yet Efficient Model for Disaster Related Data Detection
Sanjana V Herur, M Shalini, Vanshika Jain, H. R. Mamatha · 2023
Emergency scenarios pose a great challenge to impacted individuals as well as emergency response teams because they are frequently marked by unpredictability, urgency, and the need for quick action. Obtaining timely and precise information is crucial in these situations to guarantee the security and welfare of individuals experiencing difficulties. Social media platforms have become indispensable resources for emergency response and management. Among these, Twitter is essential for sharing information in real time during crisis situations because of its distinct features. Because Twitter is so open and dynamic, people may ask for assistance, discuss their experiences, and give updates on what's going on in their local areas. The platform is an essential instrument for emergency communication because of its large user base and accessibility. Therefore, Twitter, with its widespread reach and real-time updates, offers a valuable tool for identifying and responding to these requests for assistance. Not only can Twitter help identify these tweets about emergencies, but it can also help with different kinds of analyses to improve response times for disasters. These studies may involve figuring out how locals can help the rescue teams, gauging the amount of people eager to help with disaster recovery, and assisting in the rapid dissemination of information about the crisis. This study presents a simple yet efficient model for classifying tweets and assessing how relevant they are to disaster-related events. To produce text embeddings for tweets, our model makes use of Universal Sentence Encodings (USE). We determine the best classifier for binary tweet classification in this scenario through methodical experimentation accounting simplicity. According to our testing, the model effectively distinguishes between tweets that are and that are not related to disaster occurrences. It accomplishes a high degree of precision and accuracy, offering a dependable automatic tweet classification solution. Moreover, its flexibility guarantees steady performance in a range of crisis situations.