Neuro-Symbolic Entity Alignment via Variational Inference
Shengyuan Chen, Zheng Yuan, Qinggang Zhang, Wen Hua, Jiannong Cao, Xiao Huang
TL;DR
NeuSymEA is a neuro-symbolic framework for entity alignment that combines probabilistic modeling, neural inference, and symbolic reasoning to improve accuracy and robustness in knowledge graph merging tasks.
Contribution
It introduces a unified neuro-symbolic reasoning framework using variational EM for entity alignment, integrating neural and symbolic methods for better interpretability and uncertainty handling.
Findings
Achieves 7.6% improvement in hit@1 on DBP15K_ZH-EN.
Demonstrates robustness with only 1% seed pairs.
Outperforms strong baselines in low-resource settings.
Abstract
Entity alignment (EA) aims to merge two knowledge graphs (KGs) by identifying equivalent entity pairs. Existing methods can be categorized into symbolic and neural models. Symbolic models, while precise, struggle with substructure heterogeneity and sparsity, whereas neural models, although effective, generally lack interpretability and cannot handle uncertainty. We propose NeuSymEA, a unified neuro-symbolic reasoning framework that combines the strengths of both methods to fully exploit the cross-KG structural pattern for robust entity alignment. NeuSymEA models the joint probability of all possible pairs' truth scores in a Markov random field, regulated by a set of rules, and optimizes it with the variational EM algorithm. In the E-step, a neural model parameterizes the truth score distributions and infers missing alignments. In the M-step, the rule weights are updated based on the…
Peer Reviews
Decision·Submitted to ICLR 2025
1. The model achieves state-of-the-art results on entity alignment benchmark datasets, significantly improving alignment effectiveness. This validates the efficiency of the neuro-symbolic fusion approach and highlights its potential for broader applications in knowledge graph integration. 2. The inclusion of an explainer component provides rule-based interpretations for entity alignments, which is a major advantage for applications that require transparency and accountability, such as medical kn
1. The descriptions of the variational EM algorithm and inference steps are dense and complex. For readers unfamiliar with probabilistic modeling, these sections may be difficult to understand. Simplifying the explanations or including more illustrative diagrams could enhance readability and comprehension. 2. While the paper claims efficiency improvements, it lacks a thorough complexity analysis. Specifically, the impact of rule length and dataset size on runtime and memory usage should be expli
The NeuSymEA framework combines the interpretability of symbolic models with the high recall rate of neural models, optimizing entity alignment through a variational EM framework. This integration enables the model to effectively handle complex entity alignment tasks while balancing the strengths and weaknesses of both models within a unified framework.
The symbolic and neural models in the paper are integrated through the EM framework, but the trade-offs between the two in the optimization process are not explored in depth, especially in high-dimensional datasets. As a result, symbolic reasoning may still face efficiency issues, while the neural model tends to rely heavily on sparse data. Therefore, the handling strategies in these situations require further discussion.
- **Combined strengths**. The work integrates symbolic and neural models, leveraging the precision of symbolic reasoning and the high recall of neural embeddings to improve entity alignment. - **Good performance**. The proposed framework shows superior performance and robustness across benchmark datasets, outperforming existing methods. - **Interpretable results**. The proposed framework can provide interpretable results through a path-ranking-based explainer, enhancing the interpretability of
- **Unclear definition of rules**. According to Eq. (1), the rules used in the work are not horn rules. Instead, they are path pairs from anchor pairs. So, it would be better to provide a clear definition of the used rules. - **Increased complexity**. The integration of symbolic and neural models with a variational EM algorithm may be complex. No discussions or experiments are provided to analyze the complexity. The framework may still face challenges with extremely large knowledge graphs due t
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsTopic Modeling
MethodsSparse Evolutionary Training