High priority tweet detection and summarization in natural disasters

Abstract

Nowadays, Twitter is the most used microblog service and valuable source in terms of instant information. Making the intervention to the right place as soon as possible during natural disasters has great importance with regard to human life. A new system is designed and implemented that aiming to provide correct information source to help units by detecting and summarizing high priority tweets which are posted during natural disaster. To evaluate the success of the system, a dataset is created from collected tweets that posted after natural disaster and divided into two classes that carrying valuable information such as wounded and damage state are high priority and the others are low priority. Firstly, tweets are pre processed after that, classifying is made using by SVM method to detect the tweet's priority. High priority tweets are summarized using by Hybrid TF-IDF method and representing high priority tweets as best are selected.