Temporal Knowledge Graph Reasoning with Historical Contrastive Learning

TL;DR

This paper introduces CENET, a novel temporal knowledge graph reasoning model that leverages historical contrastive learning to improve event forecasting, especially for entities with limited historical data.

Contribution

The paper proposes a new training framework using contrastive learning to distinguish between historical and non-historical dependencies in temporal knowledge graphs.

Findings

CENET outperforms existing methods on five benchmark graphs.

Achieves at least 8.3% relative improvement in Hits@1.

Effectively models both historical and non-historical dependencies.

Abstract

Temporal knowledge graph, serving as an effective way to store and model dynamic relations, shows promising prospects in event forecasting. However, most temporal knowledge graph reasoning methods are highly dependent on the recurrence or periodicity of events, which brings challenges to inferring future events related to entities that lack historical interaction. In fact, the current moment is often the combined effect of a small part of historical information and those unobserved underlying factors. To this end, we propose a new event forecasting model called Contrastive Event Network (CENET), based on a novel training framework of historical contrastive learning. CENET learns both the historical and non-historical dependency to distinguish the most potential entities that can best match the given query. Simultaneously, it trains representations of queries to investigate whether the current moment depends more on historical or non-historical events by launching contrastive learning. The representations further help train a binary classifier whose output is a boolean mask to indicate related entities in the search space. During the inference process, CENET employs a mask-based strategy to generate the final results. We evaluate our proposed model on five benchmark graphs. The results demonstrate that CENET significantly outperforms all existing methods in most metrics, achieving at least $8.3\%$ relative improvement of Hits@1 over previous state-of-the-art baselines on event-based datasets.