A Survey on Temporal Knowledge Graph: Representation Learning and Applications

Li Cai,Xin Mao,Yuhao Zhou,Zhaoguang Long,Changxu Wu,Man Lan

2024-03-03

Abstract:Knowledge graphs have garnered significant research attention and are widely used to enhance downstream applications. However, most current studies mainly focus on static knowledge graphs, whose facts do not change with time, and disregard their dynamic evolution over time. As a result, temporal knowledge graphs have attracted more attention because a large amount of structured knowledge exists only within a specific period. Knowledge graph representation learning aims to learn low-dimensional vector embeddings for entities and relations in a knowledge graph. The representation learning of temporal knowledge graphs incorporates time information into the standard knowledge graph framework and can model the dynamics of entities and relations over time. In this paper, we conduct a comprehensive survey of temporal knowledge graph representation learning and its applications. We begin with an introduction to the definitions, datasets, and evaluation metrics for temporal knowledge graph representation learning. Next, we propose a taxonomy based on the core technologies of temporal knowledge graph representation learning methods, and provide an in-depth analysis of different methods in each category. Finally, we present various downstream applications related to the temporal knowledge graphs. In the end, we conclude the paper and have an outlook on the future research directions in this area.

Artificial Intelligence,Computation and Language

What problem does this paper attempt to address?

### Problems the Paper Attempts to Solve

The paper primarily focuses on **Temporal Knowledge Graphs (TKGs) representation learning and its applications**. Specifically:

1. **Background and Motivation**:
- Most current knowledge graph research mainly concentrates on static knowledge graphs, neglecting temporal information, which leads to an inability to capture the changes in entities and relationships over time.
- Temporal knowledge graphs are gaining increasing attention because they can contain structured knowledge within specific time periods.

2. **Core Contributions**:
- Conducted a comprehensive survey of existing temporal knowledge graph representation learning methods and proposed a new classification system.
- Categorized temporal knowledge graph representation learning methods into 10 categories, providing a detailed analysis of the key components, advantages, and disadvantages of each method.
- Introduced the applications of temporal knowledge graphs in time-aware knowledge reasoning, entity alignment between temporal knowledge graphs, and question answering based on temporal knowledge graphs.
- Summarized the current research progress and pointed out future research directions, including scalability, interpretability, information fusion, and integration with large language models.

3. **Specific Content**:
- Defined the basic concepts of temporal knowledge graphs and their representation learning, describing related datasets and evaluation metrics.
- Proposed various representation learning methods, such as transformation-based methods, decomposition-based methods, graph neural network-based methods, capsule network-based methods, etc., and discussed the characteristics of each method in detail.
- Explored application cases of temporal knowledge graphs in different fields, demonstrating their potential value in practical tasks.

In summary, the paper aims to promote further development in the field of temporal knowledge graph representation learning by systematically organizing and analyzing the methods and techniques, and to provide guidance for subsequent research.