A Sentiment Analysis Model for Terrorist Attacks Reviews on Twitter

Abstract

Twitter is considered as one of the famous microblogs that attract politicians and individuals to express their views on political, economic and social issues. The phenomenon of terrorist operations is one of the largest security and economic problem facing the world in recent years. Twitter users' comments on terrorism issues are important to understand users' sentiment about terrorist events. Sentiment analysis is a field of research for understanding and extracting users’ views. In this paper, we propose a model for automatically classifying users’ reviews on Twitter after occurrence of a terrorist attack, the model is built using lexicon and machine learning approaches. Lexicon approach is used to create labelled training dataset while machine learning approach was used to build the model. Scores of some domain related words were neutralized to avoid their negative effect. Features were selected based on PoS. Majority voting between NB, SVM and LR machine learning classification algorithms was applied. The performance of classification algorithms was measured using accuracy and F1 scores. The results obtained are compared to identify the best classification algorithm for features selection. Result show that our model achieved 94.8% accuracy with 95.9% F1 score.