基于关系时间嵌入的时间知识表示学习

Temporal Knowledge Representation Learning based on Relational Time Embedding

本文提出一种基于关系感知的时间嵌入RTE,扩展了现有的静态知识表示学习方法.RTE学习了关系感知的时间特征,并通过一种简单有效的融合机制耦合静态特征和时间特征.RTE是方法无关的,本文将RTE与DistMult和SimplE结合,在基准数据集上取得了先进的实验结果.关系感知时间嵌入RTE简单高效,可扩展现有的知识表示学习方法,为时间知识表示提供有益参考.

This article introduced a relationship-aware time embedding named RTE,then extended static knowledge representation learning methods with RTE.RTE,which was method-independent,learned the temporal features of relationship perception and coupled the static features with the temporal features through a simple and effective fusion mechanism.This article combined RTE with DistMult and SimplE to get the most advanced experimental results on a baseline data set.The proposed relational time-aware RTE is simple and efficient,extending the existing knowledge representation learning methods and providing a valuable reference for time knowledge representation.