Leveraging Multi-Source Weak Social Supervision for Early Detection of Fake News
- Kai Shu ,
- Guoqing Zheng ,
- Yichuan Li ,
- Subhabrata (Subho) Mukherjee ,
- Ahmed Awadallah ,
- Scott Ruston ,
- Huan Liu
European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML-PKDD 2020) |
Social media has greatly enabled people to participate in online activities at an unprecedented rate. However, this unrestricted access also exacerbates the spread of misinformation and fake news online which might cause confusion and chaos unless being detected early for its mitigation. Given the rapidly evolving nature of news events and the limited amount of annotated data, state-of-the-art systems on fake news detection face challenges due to the lack of large numbers of annotated training instances that are hard to come by for early detection. In this work, we exploit multiple weak signals from different sources given by user and content engagements (referred to as weak social supervision), and their complementary utilities to detect fake news. We jointly leverage limited amount of clean data along with weak signals from social engagements to train deep neural networks in a meta-learning framework to estimate the quality of different weak instances. Experiments on real-world datasets demonstrate that the proposed framework outperforms state-of-the-art baselines for early detection of fake news without using any user engagements at prediction time.
Publication Downloads
Multi-source Weak Social Supervision for Fake News Detection (MWSS)
May 24, 2021
This repository contains code for fake news detection with Multi-source Weak Social Supervision (MWSS), published at ECML-PKDD 2020. Social media has greatly enabled people to participate in online activities at an unprecedented rate. However, this unrestricted access also exacerbates the spread of misinformation and fake news online which might cause confusion and chaos unless being detected early for its mitigation. Given the rapidly evolving nature of news events and the limited amount of annotated data, state-of-the-art systems on fake news detection face challenges due to the lack of large numbers of annotated training instances that are hard to come by for early detection. In this work, we exploit multiple weak signals from different sources given by user and content engagements (referred to as weak social supervision), and their complementary utilities to detect fake news. We jointly leverage limited amount of clean data along with weak signals from social engagements to train deep neural networks in a meta-learning framework to estimate the quality of different weak instances. Experiments on real-world datasets demonstrate that the proposed framework outperforms state-of-the-art baselines for early detection of fake news without using any user engagements at prediction time.