Toward Degree Bias in Embedding-Based Knowledge Graph Completion
arXiv:2302.05044 · doi:10.1145/3543507.3583544
Abstract
A fundamental task for knowledge graphs (KGs) is knowledge graph completion (KGC). It aims to predict unseen edges by learning representations for all the entities and relations in a KG. A common concern when learning representations on traditional graphs is degree bias. It can affect graph algorithms by learning poor representations for lower-degree nodes, often leading to low performance on such nodes. However, there has been limited research on whether there exists degree bias for embedding-based KGC and how such bias affects the performance of KGC. In this paper, we validate the existence of degree bias in embedding-based KGC and identify the key factor to degree bias. We then introduce a novel data augmentation method, KG-Mixup, to generate synthetic triples to mitigate such bias. Extensive experiments have demonstrated that our method can improve various embedding-based KGC methods and outperform other methods tackling the bias problem on multiple benchmark datasets.
arXiv commentWWW'23
References in corpus (14)
- Adam: A Method for Stochastic Optimization
- SMOTE: Synthetic Minority Over-sampling Technique
- mixup: Beyond Empirical Risk Minimization
- On Calibration of Modern Neural Networks
- Unifying Knowledge Graph Learning and Recommendation: Towards a Better Understanding of User Preferences
- Knowledge Graph Embedding for Link Prediction: A Comparative Analysis
- GraphSMOTE: Imbalanced Node Classification on Graphs with Graph Neural Networks
- Balanced Meta-Softmax for Long-Tailed Visual Recognition
- Neural Bellman-Ford Networks: A General Graph Neural Network Framework for Link Prediction
- Investigating and Mitigating Degree-Related Biases in Graph Convolutional Networks
- Understanding the Performance of Knowledge Graph Embeddings in Drug Discovery
- Graph Data Augmentation for Graph Machine Learning: A Survey
- Positive-Unlabeled Learning with Adversarial Data Augmentation for Knowledge Graph Completion
- How Does Knowledge Graph Embedding Extrapolate to Unseen Data: A Semantic Evidence View