A Survey on Temporal Knowledge Graph Embedding

Wang TianXiang

Keywords: Temporal knowledge graph embedding, temporal information, translational distance models, semantic matching models, downstream tasks

Abstract: Knowledge Graph (KG) embedding has emerged as an active area of research. Recently, we note that KG beliefs are not universally true, as they tend to be valid only in a specific time period. However, most algorithms for KG embedding have been designed for static data, which lead to low efficiency and high error rate. Therefore, modeling dynamically-evolving, multi-relational graph data has received a surge of interests with the rapid growth of heterogeneous event data. And recent research has focused on temporal knowledge graph and its temporal information. In this survey, we first describe the standard definition of temporal knowledge graph. And then we provide a comprehensive review on temporal knowledge graph embedding covering overall research topics about temporal information and embedding models. We also introduce some prominent downstream tasks with the widely used data sets and evaluation protocol. Finally, we summarize recent challenges and highlight directions to facilitate future research.

Subtrack: Knowledge Graphs (understanding, creating, and exploiting)

First Author Is Student: Yes