Do Temporal Knowledge Graph Embedding Models Learn or Memorize Shortcuts?

Jiaxin Pan, Mojtaba Nayyeri, Yinan Li, Steffen Staab

Keywords: Temporal Knowledge Graph, Temporal Knowledge Graph Datasets, Evaluation of Existing Methods

Abstract:

Temporal Knowledge Graph Embedding models predict missing facts in temporal knowledge graphs.
Previous work on static knowledge graph embedding models has revealed that KGE models utilize shortcuts in test set leakage to achieve high performance.
In this work, we show that a similar test set leakage problem exists in widely used temporal knowledge graph datasets ICEWS14 and ICEWS05-15. We propose a naive rule-based model that can achieve start-of-the-art results on both datasets without a deep-learning process. Following this consideration, we construct two more challenging datasets for the evaluation of TKGEs.

Format: Short paper, up to 4 pages.

Submission Number: 49