During natural disasters and emergencies, social media platforms, especially Twitter, serve as an important channel for real-time information sharing. However, not every tweet, even those including a disaster-related keyword, actually refers to a real disaster event, for example: “I am on fire for this exam” v.s. “The building is on fire”. Thus, distinguishing between tweets which describe real disasters and tweets which are unrelated is an important Our core problem is thus: using machine learning method to accurately classify whether a tweet is about a real disaster or not.
This classification can be used for helping quick response when natural disasters happen
In order to use the code, you may include the data files and Model files in local path, they can be accessed on Kaggle and HuggingFace.