Understanding and Guiding Weakly Supervised Entity Alignment with Potential Isomorphism Propagation

TL;DR

This work addresses weakly supervised entity alignment by reframing it as a propagation problem over cross-graph similarities. It introduces a theoretical foundation showing that potentially aligned entities lie in isomorphic subgraphs and proposes the potential isomorphism propagation operator to bridge intra- and inter-graph information. Building on this, PipEA couples an isomorphism-aware propagation mechanism with a refinement scheme and scalable matrix factorization to produce robust cross-graph similarities, improving one-to-one alignment with minimal seeds. Empirical results on OpenEA v2 across cross-lingual and mono-lingual benchmarks demonstrate state-of-the-art gains, and ablations validate the contributions of each component. Overall, the work advances both the theory and practice of aggregation-based weakly supervised EA by enabling effective neighborhood propagation across heterogeneous knowledge graphs.

Abstract

Weakly Supervised Entity Alignment (EA) is the task of identifying equivalent entities across diverse knowledge graphs (KGs) using only a limited number of seed alignments. Despite substantial advances in aggregation-based weakly supervised EA, the underlying mechanisms in this setting remain unexplored. In this paper, we present a propagation perspective to analyze weakly supervised EA and explain the existing aggregation-based EA models. Our theoretical analysis reveals that these models essentially seek propagation operators for pairwise entity similarities. We further prove that, despite the structural heterogeneity of different KGs, the potentially aligned entities within aggregation-based EA models have isomorphic subgraphs, which is the core premise of EA but has not been investigated. Leveraging this insight, we introduce a potential isomorphism propagation operator to enhance the propagation of neighborhood information across KGs. We develop a general EA framework, PipEA, incorporating this operator to improve the accuracy of every type of aggregation-based model without altering the learning process. Extensive experiments substantiate our theoretical findings and demonstrate PipEA's significant performance gains over state-of-the-art weakly supervised EA methods. Our work not only advances the field but also enhances our comprehension of aggregation-based weakly supervised EA.

Figures (4)

• Figure 1: (1) Propagation gap in weakly supervised settings and (3) our potential isomorphism propagation.
• Figure 2: PipEA initiates with pairwise similarities from aggregation-based models, employs potential isomorphism propagation to generate new similarities, and introduces a refinement scheme to effectively integrate them.
• Figure 3: Similarity matrix of our method and PEEA encoder on sub-graph from 15KEN-DE.
• Figure 4: Hyper-parameter experiments of our method.