Detection and summarization of the high priority tweets after natural disasters

Twitter, being one of the most widely used micro-blog services, has been one of the best sources of instant data. It has a great significance in terms of saving lives during natural disasters by means of determining the right place for taking prompt lifesaving efforts. In this study, a system that identifies the tweets of a higher priority written immediately after natural disasters and classifies them in order to further send the right data to first responders or aid units, has been designed and realized. To evaluate how the system works, a database has been arranged containing tweets written after natural disasters and classified into two categories--tweets containing valuable information about injuries and damage being marked as those of a high priority and the other tweets marked as those of a low priority. First of all, the tweets are pre-processed to clean out the noise and evaluate the classifiers in a more successful way. Then, using the classification by means of the Support Vector Machine method, the tweets are decided on whether they are of a high priority or not. Finally, summarized by the Hybrid TF-IDF method, the high priority tweets that best represent the cluster have been selected.