Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Rumor Detection on Social Media with Temporal Propagation Structure Optimization

Xingyu Peng, Junran Wu, Ruomei Liu, Ke Xu

TL;DR

The paper addresses rumor veracity prediction by integrating temporal dynamics into propagation-structure modeling. It constructs time-weighted propagation trees and derives coding trees via structural-entropy minimization, then learns rumor representations with a CT-RvNN that propagates information bottom-up and reads out a hierarchical embedding. Empirical results on PHEME and Rumoreval show CT-RvNN outperforms state-of-the-art methods while using fewer parameters, and ablations confirm the value of temporal weights and coding-tree optimization. The approach offers practical benefits for early detection and robust rumor classification in noisy propagation-graphs.

Abstract

Traditional methods for detecting rumors on social media primarily focus on analyzing textual content, often struggling to capture the complexity of online interactions. Recent research has shifted towards leveraging graph neural networks to model the hierarchical conversation structure that emerges during rumor propagation. However, these methods tend to overlook the temporal aspect of rumor propagation and may disregard potential noise within the propagation structure. In this paper, we propose a novel approach that incorporates temporal information by constructing a weighted propagation tree, where the weight of each edge represents the time interval between connected posts. Drawing upon the theory of structural entropy, we transform this tree into a coding tree. This transformation aims to preserve the essential structure of rumor propagation while reducing noise. Finally, we introduce a recursive neural network to learn from the coding tree for rumor veracity prediction. Experimental results on two common datasets demonstrate the superiority of our approach.

Rumor Detection on Social Media with Temporal Propagation Structure Optimization

TL;DR

The paper addresses rumor veracity prediction by integrating temporal dynamics into propagation-structure modeling. It constructs time-weighted propagation trees and derives coding trees via structural-entropy minimization, then learns rumor representations with a CT-RvNN that propagates information bottom-up and reads out a hierarchical embedding. Empirical results on PHEME and Rumoreval show CT-RvNN outperforms state-of-the-art methods while using fewer parameters, and ablations confirm the value of temporal weights and coding-tree optimization. The approach offers practical benefits for early detection and robust rumor classification in noisy propagation-graphs.

Abstract

Traditional methods for detecting rumors on social media primarily focus on analyzing textual content, often struggling to capture the complexity of online interactions. Recent research has shifted towards leveraging graph neural networks to model the hierarchical conversation structure that emerges during rumor propagation. However, these methods tend to overlook the temporal aspect of rumor propagation and may disregard potential noise within the propagation structure. In this paper, we propose a novel approach that incorporates temporal information by constructing a weighted propagation tree, where the weight of each edge represents the time interval between connected posts. Drawing upon the theory of structural entropy, we transform this tree into a coding tree. This transformation aims to preserve the essential structure of rumor propagation while reducing noise. Finally, we introduce a recursive neural network to learn from the coding tree for rumor veracity prediction. Experimental results on two common datasets demonstrate the superiority of our approach.

Paper Structure

This paper contains 37 sections, 1 theorem, 15 equations, 10 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Rumor Detection.
  4. Structural Entropy.
  5. Preliminaries
  6. Problem Statement
  7. Structural Entropy
  8. Methodology
  9. Time-weighted Propagation Tree Construction
  10. Coding Tree Construction
  11. Rumor Representation Learning
  12. Leaf Node Encoding.
  13. Tree Positional Encoding.
  14. Bottom-Up Message Passing.
  15. Tree Representation Readout.
  16. ...and 22 more sections

Key Result

Proposition 1

For two nodes $v_j$ and $v_k$ in a coding tree $T$ such that $v_j$ is the ancestor of $v_k$, we have $\mathcal{H}^T\left(G\right) = \mathcal{H}^{T.\textit{pad}\,(v_k)}\left(G\right)$.

Figures (10)

  • Figure 1: Example of a rumor propagation tree related to the Ferguson event.
  • Figure 2: The Empirical Cumulative Distribution Function (ECDF) plots of the time delay distributions since the initial claim was posted for posts responding to true, false, and unverified rumors in the Germanwings crash event from the PHEME zubiaga2016analysing dataset.
  • Figure 3: Overview of our approach.
  • Figure 4: Performance analysis results on the PHEME and Rumoreval datasets in terms of coding tree height.
  • Figure 5: Early rumor detection performance analysis results for four distinct events from the PHEME dataset.
  • ...and 5 more figures

Theorems & Definitions (6)

  • Definition 1
  • Definition 2
  • Definition 3
  • Proposition 1
  • proof
  • proof