Tension detection in online communities

Abstract

The use of online platforms to express views and opinions is creating an enormous amount of data from which relevant information may be extracted and utilized in a number of ways. This paper explores the use of this vast and readily available information for the cause of ensuring peace and harmony in the society. The work collects and analyzes the available online data to detect if tension in the society is aggravated due to a particular issue. The data set, consisting of tweets and comments on the news of an issue selected by the user, is given as an input to two types of classifiers, namely Naive Bayes and Support Vector Machine. The classifiers divide the documents into three classes, namely high tension low tension and moderate tension. These classes are given weights to identify and plot tension vs. time graphs. This system may be used by the police, social service agencies, policy makers, etc to determine the opinion of the masses on various issues and its effect on society.