A Tool for Fake News Detection

The word post-truth was considered by Oxford Dictionaries Word of the Year 2016. The word is an adjective relating to or denoting circumstances in which objective facts are less influential in shaping public opinion than appeals to emotion and personal belief. This leads to misinformation and problems in society. Hence, it is important to make effort to detect these facts and prevent them from spreading. In this paper we propose machine learning techniques, in particular supervised learning, for fake news detection. More precisely, we used a dataset of fake and real news to train a machine learning model using Scikit-learn library in Python. We extracted features from the dataset using text representation models like Bag-of-Words, Term Frequency-Inverse Document Frequency (TF-IDF) and Bi-gram frequency. We tested two classification approaches, namely probabilistic classification and linear classification on the title and the content, checking if it is clickbait/nonclickbait, respectively fake/real. The outcome of our experiments was that the linear classification works the best with the TF-IDF model in the process of content classification. The Bi-gram frequency model gave the lowest accuracy for title classification in comparison with Bag-of-Words and TF-IDF.