Learning with Dual-level Noisy Correspondence for Multi-modal Entity Alignment

2025-10-20

AI Summary

This paper addresses the dual-level noise correspondence (DNC) problem prevalent in multimodal knowledge graphs (MMKGs), wherein intra-entity attribute noise and cross-graph entity/attribute alignment noise coexist. To tackle this challenge, we propose RULE—the first systematic modeling framework for DNC. RULE comprises four synergistic components: dual-branch reliability estimation, reliability-aware attribute fusion, cross-graph discrepancy elimination, and logic-driven attribute association reasoning. Its core innovations lie in a reliability-aware hierarchical noise modeling mechanism and a joint cross-modal and cross-graph reasoning paradigm. Extensive experiments on five benchmark datasets demonstrate that RULE consistently outperforms seven state-of-the-art methods, achieving absolute improvements of 4.2–9.7% in entity alignment accuracy under high-noise conditions. These results validate RULE’s robustness and generalizability in handling complex, realistic MMKG noise scenarios.

Technology Category

• Data Mining & Knowledge Management: Rule Mining & Pattern Mining
• Knowledge Representation and Reasoning: Knowledge Acquisition
• Machine Learning: Graph-based Machine Learning

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Multi-modal entity alignment (MMEA) aims to identify equivalent entities across heterogeneous multi-modal knowledge graphs (MMKGs), where each entity is described by attributes from various modalities. Existing methods typically assume that both intra-entity and inter-graph correspondences are faultless, which is often violated in real-world MMKGs due to the reliance on expert annotations. In this paper, we reveal and study a highly practical yet under-explored problem in MMEA, termed Dual-level Noisy Correspondence (DNC). DNC refers to misalignments in both intra-entity (entity-attribute) and inter-graph (entity-entity and attribute-attribute) correspondences. To address the DNC problem, we propose a robust MMEA framework termed RULE. RULE first estimates the reliability of both intra-entity and inter-graph correspondences via a dedicated two-fold principle. Leveraging the estimated reliabilities, RULE mitigates the negative impact of intra-entity noise during attribute fusion and prevents overfitting to noisy inter-graph correspondences during inter-graph discrepancy elimination. Beyond the training-time designs, RULE further incorporates a correspondence reasoning module that uncovers the underlying attribute-attribute connection across graphs, guaranteeing more accurate equivalent entity identification. Extensive experiments on five benchmarks verify the effectiveness of our method against the DNC compared with seven state-of-the-art methods.The code is available at href{https://github.com/XLearning-SCU/RULE}{XLearning-SCU/RULE}

Problem

• Addresses noisy entity-attribute correspondences in multi-modal knowledge graphs
• Mitigates inter-graph misalignments between entities and attributes
• Improves robustness against dual-level noise for entity alignment

Innovation

• Estimates correspondence reliability via two-fold principle
• Mitigates intra-entity noise during attribute fusion
• Incorporates cross-graph correspondence reasoning module