[Paper Review] FAKEDETECTOR: Effective Fake News Detection with Deep Diffusive Neural Network

Jiawei Zhang, Bowen Dong+1 more|arXiv (Cornell University)|May 22, 2018

Misinformation and Its Impacts68 references52 citations

TL;DR

This paper presents FakeDetector, a deep diffusive network with a hybrid feature learning unit to jointly infer credibility for fake news articles, creators, and subjects using textual and relational signals in a news-augmented heterogeneous social network.

ABSTRACT

In recent years, due to the booming development of online social networks, fake news for various commercial and political purposes has been appearing in large numbers and widespread in the online world. With deceptive words, online social network users can get infected by these online fake news easily, which has brought about tremendous effects on the offline society already. An important goal in improving the trustworthiness of information in online social networks is to identify the fake news timely. This paper aims at investigating the principles, methodologies and algorithms for detecting fake news articles, creators and subjects from online social networks and evaluating the corresponding performance. This paper addresses the challenges introduced by the unknown characteristics of fake news and diverse connections among news articles, creators and subjects. This paper introduces a novel automatic fake news credibility inference model, namely FAKEDETECTOR. Based on a set of explicit and latent features extracted from the textual information, FAKEDETECTOR builds a deep diffusive network model to learn the representations of news articles, creators and subjects simultaneously. Extensive experiments have been done on a real-world fake news dataset to compare FAKEDETECTOR with several state-of-the-art models, and the experimental results have demonstrated the effectiveness of the proposed model.

Motivation & Objective

• Address the problem of fake news detection in online social networks with unknown characteristics and diverse article-creator-subject connections.
• Jointly infer credibility scores for news articles, creators, and subjects by leveraging textual signals and network relations.
• Develop a deep learning model that fuses heterogeneous information through a novel diffusive architecture.

Proposed method

• Introduce a hybrid feature learning unit (HFLU) to extract explicit textual features and latent features via RNN (GRU) on article, creator, and subject text.
• Propose a deep diffusive network with gated diffusive units (GDU) to fuse information across articles, creators, and subjects.
• Define an end-to-end objective with cross-entropy losses for articles, creators, and subjects plus regularization.
• Train the model with back-propagation on PolitiFact data and predict credibility labels for all node types.

Experimental results

Research questions

• RQ1 How can explicit and latent textual features from articles, creators, and subjects enhance fake news credibility inference?
• RQ2 Can a deep diffusive network effectively fuse heterogeneous relations (authorship and article-subject links) to improve detection performance across articles, creators, and subjects?
• RQ3 What is the impact of jointly modeling articles, creators, and subjects on credibility prediction in PolitiFact data?

Key findings

• The framework achieves credibility inference for news articles, creators, and subjects using a unified model on real-world data.
• The PolitiFact dataset comprises 14,055 articles, 3,634 creators, and 152 subjects with 48,756 article-subject links and 14,055 creator-article links.
• A deep diffusive network with gated diffusive units leverages both explicit word features and latent textual representations.
• The model incorporates a forget gate for subject inputs and an adjust gate for creator inputs to manage cross-type information flow.