Twitter Based Flood Rescue System

During a disaster-related event, users post different types of information like personal opinions, sentiments, and situational updates in huge volume. Information regarding situational updates will be useful for helping agencies. Our objective is to facilitate efforts required for relief operations using the real-time Twitter streaming data. Thus, we are interested in a system to identify such information which helps the emergency management agencies to provide help to victims. The proposed approach performs five essential tasks: (i) identifying most relevant tweets based on semantic similarity calculation between the tweet and query terms by utilizing the Word2Vec word embedding model, (ii) extracting location information from relevant tweets to know where is help required or provided, (iii) filtering those tweets that do not contain contact information (location or contact number), (iv) ranking of tweets based on the emergency of the situation to provide help, and (v) visualization of top-ranked tweets on Google Map.