MEMEWEAVER: Inter-Meme Graph Reasoning for Sexism and Misogyny Detection

2026-01-13

AI Summary

Existing content moderation approaches struggle to detect gender-based discrimination and misogynistic content embedded in memes, as they often overlook community interactions and social dynamics. This work proposes an end-to-end trainable multimodal framework that models semantic and social relationships among memes through a novel cross-meme graph reasoning mechanism, overcoming the limitations of heuristic graph construction and instance-level reasoning. By integrating visual and textual information, the method employs a learnable graph neural network to construct semantically meaningful graph structures. Evaluated on the MAMI and EXIST benchmarks, the approach outperforms state-of-the-art methods with faster convergence and yields graph structures that exhibit strong semantic interpretability.

Technology Category

• Machine Learning: Multimodal Learning
• Computer Vision: Multi-modal Vision
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Women are twice as likely as men to face online harassment due to their gender. Despite recent advances in multimodal content moderation, most approaches still overlook the social dynamics behind this phenomenon, where perpetrators reinforce prejudices and group identity within like-minded communities. Graph-based methods offer a promising way to capture such interactions, yet existing solutions remain limited by heuristic graph construction, shallow modality fusion, and instance-level reasoning. In this work, we present MemeWeaver, an end-to-end trainable multimodal framework for detecting sexism and misogyny through a novel inter-meme graph reasoning mechanism. We systematically evaluate multiple visual--textual fusion strategies and show that our approach consistently outperforms state-of-the-art baselines on the MAMI and EXIST benchmarks, while achieving faster training convergence. Further analyses reveal that the learned graph structure captures semantically meaningful patterns, offering valuable insights into the relational nature of online hate.

Problem

• sexism detection
• misogyny detection
• inter-meme graph
• multimodal reasoning
• online harassment

Innovation

• inter-meme graph reasoning
• multimodal fusion
• sexism detection
• graph-based reasoning
• online hate analysis