Hoax Detection On Twitter

Abstract—In this paper, we present our work on hoax detection on a collection of Tweets. We tackle the hoax detection as a text classification problem, with Doc2Vec as the text representation method and SVM as the classifier. We collected and annotated 5000 Tweets that consist of 2500 hoax Tweets and 2500 truth Tweets. The experimental results show that the accuracy of our proposed hoax detection on Tweets is 93.4%. Index Terms—Hoax Detection, Natural Language Preprocessing