3HAN: A Deep Neural Network for Fake News Detection Sneha Singhania (B ) , Nigel Fernandez, and Shrisha Rao International Institute of Information Technology - Bangalore, Bangalore, India {sneha.a,nigelsteven.fernandez}@iiitb.org, shrao@ieee.org Abstract. The rapid spread of fake news is a serious problem calling for AI solutions. We employ a deep learning based automated detector through a three level hierarchical attention network (3HAN) for fast, accurate detection of fake news. 3HAN has three levels, one each for words, sentences, and the headline, and constructs a news vector: an effective representation of an input news article, by processing an article in an hierarchical bottom-up manner. The headline is known to be a dis- tinguishing feature of fake news, and furthermore, relatively few words and sentences in an article are more important than the rest. 3HAN gives a differential importance to parts of an article, on account of its three layers of attention. By experiments on a large real-world data set, we observe the effectiveness of 3HAN with an accuracy of 96.77%. Unlike some other deep learning models, 3HAN provides an understandable out- put through the attention weights given to different parts of an article, which can be visualized through a heatmap to enable further manual fact checking. Keywords: Fake news · Deep learning · Text representation · Attention mechanism · Text classification 1 Introduction The spread of fake news is a matter of concern due to its possible role in manip- ulating public opinion. We define fake news in line with The New York Times as a “made up story with the intention to deceive, often with monetary gain as a motive” [1]. The fake news problem is complex given its varied interpretations across demographics. We present a three level hierarchical attention network (3HAN) which creates an effective representation of a news article called news vector. A news vector can be used to classify an article by assigning a probability of being fake. Unlike other neural models which are opaque in their internal reasoning and give results that are difficult to analyze, 3HAN provides an importance score for each word and sentence of an input article based on its relevance in arriving at the output probability of that article being fake. These importance scores can be visualized S. Singhania and N. Fernandez—These authors contributed equally to this work. c  Springer International Publishing AG 2017 D. Liu et al. (Eds.): ICONIP 2017, Part II, LNCS 10635, pp. 1–10, 2017. https://doi.org/10.1007/978-3-319-70096-0_59