Fake News Detection on Social Media: A Word Embedding-Based Approach

The rapid development of social media, together with the large number of user-generated content on them, has not only connected an unprecedented number of people together to do good stuff, but also has provided convenient platforms to spread misleading pieces of information such as fake news. Existing research has attempted to leverage machine learning to automatically classify fake news. In this paper, we extend such literature by proposing an approach that utilize word embedding and Long Short-Term Memory (LSTM) neural network algorithm. Unlike existing studies, we used two publicly available datasets of news articles to evaluate the proposed model. The results demonstrated the effectiveness of our model against the baseline machine learning models with accuracy of 99% and 96% using the first and second datasets respectively. These comparatively better results and effectiveness compared to existing models demonstrate that pre-trained word embedding models play a significant role in the fake news detection. Keywor