Keywords: Knowledge Graph Completion, Temporal Knowledge Graphs, Representation Learning, Few-Shot Learning, Meta-Learning
Abstract: Most real-world knowledge graphs are characterized by a frequency distribution with a long-tail where a significant fraction of relations occurs only a handful of times. This observation has given rise to recent interest in low-shot learning methods that are able to generalize from only a few examples per relation. The existing approaches, however, are tailored to static knowledge graphs and do not easily generalize to temporal settings, where data scarcity poses even bigger problems, e.g., due to occurrence of new, previously unseen relations. We address this shortcoming by proposing a one-shot learning framework for link prediction in temporal knowledge graphs. Our proposed method employs a self-attention mechanism to effectively encode temporal interactions between entities, and a network to compute a similarity score between a given query and a (one-shot) example. Our experiments show that the proposed algorithm outperforms the state of the art baselines for two well-studied benchmarks while achieving significantly better performance for sparse relations.
Subject Areas: Knowledge Representation, Semantic Web and Search, Machine Learning
Archival Status: Archival
Supplementary Material: zip