Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Learning Complex Heterogeneous Multimodal Fake News via Social Latent Network Inference

Mingxin Li, Yuchen Zhang, Haowei Xu, Xianghua Li, Chao Gao, Zhen Wang

TL;DR

This work tackles fake news detection in complex multimodal settings by constructing a latent social network through event-based cascade inference using an improved Hawkes process, and by building a personalized heterogeneous graph over news attributes. It introduces a self-supervised multimodal content learning strategy that prunes, augments, and contrasts unimodal and cross-modal features, integrated via a multimodal loss and a graph-transformer-based fusion. The approach achieves state-of-the-art performance on benchmark datasets (e.g., FakeSV, FVC), outperforming unimodal, multimodal, and LLM-based baselines, and is shown to be robust via ablation studies and visualizations. The framework’s latent-network perspective and self-supervised feature learning offer practical advantages for scalable, cross-platform fake news detection in real-world social media environments, with strong potential for extension as a plugin to other cross-modal analyses.

Abstract

With the diversification of online social platforms, news dissemination has become increasingly complex, heterogeneous, and multimodal, making the fake news detection task more challenging and crucial. Previous works mainly focus on obtaining social relationships of news via retweets, limiting the accurate detection when real cascades are inaccessible. Given the proven assessment of the spreading influence of events, this paper proposes a method called HML (Complex Heterogeneous Multimodal Fake News Detection method via Latent Network Inference). Specifically, an improved social latent network inference strategy is designed to estimate the maximum likelihood of news influences under the same event. Meanwhile, a novel heterogeneous graph is built based on social attributes for multimodal news under different events. Further, to better aggregate the relationships among heterogeneous multimodal features, this paper proposes a self-supervised-based multimodal content learning strategy, to enhance, align, fuse and compare heterogeneous modal contents. Based above, a personalized heterogeneous graph representation learning is designed to classify fake news. Extensive experiments demonstrate that the proposed method outperforms the SOTA in real social media news datasets.

Learning Complex Heterogeneous Multimodal Fake News via Social Latent Network Inference

TL;DR

This work tackles fake news detection in complex multimodal settings by constructing a latent social network through event-based cascade inference using an improved Hawkes process, and by building a personalized heterogeneous graph over news attributes. It introduces a self-supervised multimodal content learning strategy that prunes, augments, and contrasts unimodal and cross-modal features, integrated via a multimodal loss and a graph-transformer-based fusion. The approach achieves state-of-the-art performance on benchmark datasets (e.g., FakeSV, FVC), outperforming unimodal, multimodal, and LLM-based baselines, and is shown to be robust via ablation studies and visualizations. The framework’s latent-network perspective and self-supervised feature learning offer practical advantages for scalable, cross-platform fake news detection in real-world social media environments, with strong potential for extension as a plugin to other cross-modal analyses.

Abstract

With the diversification of online social platforms, news dissemination has become increasingly complex, heterogeneous, and multimodal, making the fake news detection task more challenging and crucial. Previous works mainly focus on obtaining social relationships of news via retweets, limiting the accurate detection when real cascades are inaccessible. Given the proven assessment of the spreading influence of events, this paper proposes a method called HML (Complex Heterogeneous Multimodal Fake News Detection method via Latent Network Inference). Specifically, an improved social latent network inference strategy is designed to estimate the maximum likelihood of news influences under the same event. Meanwhile, a novel heterogeneous graph is built based on social attributes for multimodal news under different events. Further, to better aggregate the relationships among heterogeneous multimodal features, this paper proposes a self-supervised-based multimodal content learning strategy, to enhance, align, fuse and compare heterogeneous modal contents. Based above, a personalized heterogeneous graph representation learning is designed to classify fake news. Extensive experiments demonstrate that the proposed method outperforms the SOTA in real social media news datasets.
Paper Structure (35 sections, 14 equations, 6 figures, 3 tables)

This paper contains 35 sections, 14 equations, 6 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Preliminary
  3. Methodology
  4. Social Latent Network Inference (Stage 1)
  5. Event-based Cascade Inference via Hawkes Process
  6. Heterogeneous Graph Construction
  7. Self-supervised-based Multimodal Content Learning (Stage 2)
  8. Pruning and Mask
  9. Unimodal Content Augmentation
  10. Cross-modal Contrast Enhancement
  11. Multimodal Loss Integration
  12. Network-Augmented Heterogeneous Multimodal Fake News Detection Task
  13. Edge Types Combination
  14. Classification
  15. Experiments
  16. ...and 20 more sections

Figures (6)

  • Figure 1: Differences in news dissemination across social media platforms.
  • Figure 2: Illustrate of proposed event-based cascade influence. The red line represents the influence intensity of Event 1 over time $t$, which is affected by Event 2 (gray line).
  • Figure 3: Framework of HML. The whole framework is divided into three modules.
  • Figure 4: An example of the proposed Heterogeneous Graph Construction strategy in dataset FakeSV.
  • Figure 5: T-SNE Visualization of FakeSV dataset.
  • ...and 1 more figures

Theorems & Definitions (1)

  • Definition 1: Attribute Heterogeneous News Graph