Context-Aware Deep Markov Random Fields for Fake News Detection

Fake news is a serious problem, which has received considerable attention from both industry and academic communities. Over the past years, many fake detection approaches have been introduced, most of existing methods rely on either content or social context dissemination process media platforms. In this work, we propose generic model that able to take into account for identification news. Specifically, explore different aspects by using shallow deep representations. The representations are produced with word2vec doc2vec models while generated via transformer-based models. These jointly separately address four individual tasks, namely bias detection, clickbait sentiment analysis, toxicity detection. addition, make use graph convolutional neural networks mean-field layers in order exploit underlying structural information articles. That way, inherent correlation between articles leveraging their information. Experiments widely-used benchmark datasets indicate effectiveness proposed method.