Classifying Arabic Tweets Based on Credibility Using Content and User Features

TLDR: This paper utilizes content-and user-related features, and employs sentiment analysis to generate new features for the detection of fake Arabic news, and shows that the system can filter out fake news with an accuracy of 76%.

Abstract: Social Media services, such as Facebook and Twitter, have recently become a huge and continuous source of daily news. People all around the world rely heavily on news published via social media to know more about current events and activities. As a result, many users have started to exploit social media by broadcasting misleading news for financial and political purposes, which has an adverse impact on society. In this paper, we utilize machine learning to identify fake news from Arabic tweets based on a supervised classification model. Twitter content published in Arabic is very noisy with a high level of uncertainty, where little work has been accomplished to process and extract important features for classification purposes. In this paper, we utilize content-and user-related features, and employ sentiment analysis to generate new features for the detection of fake Arabic news. Sentiment analysis led to improving the accuracy of the prediction process. Among a number of machine learning algorithms used to train the classification models, four algorithms are chosen, namely Random Forest, Decision Tree, AdaBoost, and Logistic Regression. The experimental evaluation shows that our system can filter out fake news with an accuracy of 76%.