Characterizing and Countering Communal and Anti Communal Tweets During Disasters

Abstract

Various tweets shared during a disaster situation encompasses data related to current scenario and about emotions/opinions. By analyzing these communal tweets, abusive posts which targets various religiousandracial groups during natural calamities has been found. By reviewingits effects, a classifier has been developed to distinguish between communal and non-communalmessages, which shows better performance. People posting such communal tweets has been analyzed which says that most of them are posted by popular users from media, politicsand form strong correlated groups in the social network which makes it to reach higher. An event-independent classifier has been proposed whichidentifiesanti-communal tweets automatically and propose a way to counter back. A real-time service has been developed to find tweets automatically related to an emergency segregating communal and anti-communal tweets. Government and local monitoring agencies can use this system for making decisions like filtering or to promote some news.