# Early Detection of Fake News by Utilizing the Credibility of News, Publishers, and Users Based on Weakly Supervised Learning

**Authors:** Chunyuan Yuan, Qianwen Ma, Wei Zhou, Jizhong Han, Songlin Hu  
**arXiv:** [2012.04233](https://arxiv.org/abs/2012.04233)  
**Date:** 2020-12-08  
**Summary page:** https://www.summarizepaper.com/en/paper/2012.04233v2/

## Key points

- Fake news is a major concern due to its negative impact on personal reputation and public trust.
- Previous studies have focused on finding clues from news content or diffusion path but the required features of these models are often unavailable or insufficient in early detection scenarios, resulting in poor performance.
- Researchers have proposed a novel Structure-aware Multi-head Attention Network (SMAN) that combines news content, publishing and reposting relations of publishers and users to jointly optimize fake news detection and credibility prediction tasks.
- SMAN can quickly locate fake news in massive news and detect them in the early stages of dissemination by using the credibility of publishers and users as prior weakly supervised information.
- The model explicitly exploits the credibility of publishers and users for early fake news detection by combining their heterogeneous graphs with attention mechanisms.
- Experiments conducted on three real-world datasets show that SMAN can detect fake news in just four hours with an accuracy of over 91%, which is much faster than state-of-the-art models.
- Recent studies have applied various neural networks to automatically learn high-level representations for fake news detection such as RNNs, CNNs, matrix factorization and graph neural networks. However, these methods pay little attention to early detection.
- The proposed SMAN model addresses this gap by leveraging the credibility of publishers and users to facilitate early fake news detection while surpassing other state-of-the-art models on both fake news classification and early detection tasks while significantly reducing the time required for detection.
- Overall, this study provides an effective solution for detecting fake news at an early stage before it spreads widely across social media platforms.

## Summary

Fake news has become a major concern in recent years due to its negative impact on personal reputation and public trust. Previous studies have focused on finding clues from news content or diffusion path but the required features of these models are often unavailable or insufficient in early detection scenarios, resulting in poor performance. To address this issue, researchers have proposed a novel Structure-aware Multi-head Attention Network (SMAN) that combines news content, publishing and reposting relations of publishers and users to jointly optimize fake news detection and credibility prediction tasks. By using the credibility of publishers and users as prior weakly supervised information, SMAN can quickly locate fake news in massive news and detect them in the early stages of dissemination. The model explicitly exploits the credibility of publishers and users for early fake news detection by combining their heterogeneous graphs with attention mechanisms. Experiments conducted on three real-world datasets show that SMAN can detect fake news in just four hours with an accuracy of over 91%, which is much faster than state-of-the-art models. Early research on fake news detection mainly focused on designing effective features from various sources such as textual content, user profiling data and news diffusion patterns. However, these feature-based methods are time-consuming, biased and require a lot of labor to design while also being easily manipulated by users. Recent studies have applied various neural networks to automatically learn high-level representations for fake news detection such as RNNs, CNNs, matrix factorization and graph neural networks. While these methods apply more types of information for fake news detection they pay little attention to early detection. The proposed SMAN model addresses this gap by leveraging the credibility of publishers and users to facilitate early fake news detection while surpassing other state-of-the-art models on both fake news classification and early detection tasks while significantly reducing the time required for detection. Overall, this study provides an effective solution for detecting fake news at an early stage before it spreads widely across social media platforms.