Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Transferring Structure Knowledge: A New Task to Fake news Detection Towards Cold-Start Propagation

Lingwei Wei, Dou Hu, Wei Zhou, Songlin Hu

TL;DR

The paper addresses the practical challenge of detecting fake news when propagation data is unavailable for new samples (cold-start propagation). It introduces the Structure Adversarial Net (SAN), a transfer-learning framework that uses a structure discriminator and gradient-reversal to learn structure-invariant representations from propagation-rich data and apply them to content-only samples. Empirical results across PolitiFact, GossipCop, and PHEME-5 show that SAN consistently improves propagation-based detectors under general and event-aware cold-start scenarios, highlighting improved generalization and transferability. The work offers a model-agnostic approach to extend propagation-based fake news detection to real-world, propagation-sparse settings and suggests directions for further optimization and datasets.

Abstract

Many fake news detection studies have achieved promising performance by extracting effective semantic and structure features from both content and propagation trees. However, it is challenging to apply them to practical situations, especially when using the trained propagation-based models to detect news with no propagation data. Towards this scenario, we study a new task named cold-start fake news detection, which aims to detect content-only samples with missing propagation. To achieve the task, we design a simple but effective Structure Adversarial Net (SAN) framework to learn transferable features from available propagation to boost the detection of content-only samples. SAN introduces a structure discriminator to estimate dissimilarities among learned features with and without propagation, and further learns structure-invariant features to enhance the generalization of existing propagation-based methods for content-only samples. We conduct qualitative and quantitative experiments on three datasets. Results show the challenge of the new task and the effectiveness of our SAN framework.

Transferring Structure Knowledge: A New Task to Fake news Detection Towards Cold-Start Propagation

TL;DR

The paper addresses the practical challenge of detecting fake news when propagation data is unavailable for new samples (cold-start propagation). It introduces the Structure Adversarial Net (SAN), a transfer-learning framework that uses a structure discriminator and gradient-reversal to learn structure-invariant representations from propagation-rich data and apply them to content-only samples. Empirical results across PolitiFact, GossipCop, and PHEME-5 show that SAN consistently improves propagation-based detectors under general and event-aware cold-start scenarios, highlighting improved generalization and transferability. The work offers a model-agnostic approach to extend propagation-based fake news detection to real-world, propagation-sparse settings and suggests directions for further optimization and datasets.

Abstract

Many fake news detection studies have achieved promising performance by extracting effective semantic and structure features from both content and propagation trees. However, it is challenging to apply them to practical situations, especially when using the trained propagation-based models to detect news with no propagation data. Towards this scenario, we study a new task named cold-start fake news detection, which aims to detect content-only samples with missing propagation. To achieve the task, we design a simple but effective Structure Adversarial Net (SAN) framework to learn transferable features from available propagation to boost the detection of content-only samples. SAN introduces a structure discriminator to estimate dissimilarities among learned features with and without propagation, and further learns structure-invariant features to enhance the generalization of existing propagation-based methods for content-only samples. We conduct qualitative and quantitative experiments on three datasets. Results show the challenge of the new task and the effectiveness of our SAN framework.
Paper Structure (10 sections, 10 equations, 3 figures, 2 tables)

This paper contains 10 sections, 10 equations, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. A New Task: Cold-start Fake news Detection
  3. Approach
  4. Vanilla Propagation-based Approach
  5. Structure Adversarial Net Framework
  6. Experiments
  7. Experimental Setups
  8. Task Analysis
  9. Effects of SAN Framework
  10. Conclusion

Figures (3)

  • Figure 1: The architecture of SAN framework for cold-start fake news detection. A structure discriminator is introduced to predict the auxiliary structure label based on the latent representation. SAN can transfer structure knowledge learned from existing propagation trees to content-only samples.
  • Figure 2: Results on different fake news detection tasks.
  • Figure 3: Visualization of representations on PolitiFact.