|
基于时序知识图谱补全的关系预测
|
Abstract:
知识图谱是对客观世界中实体、概念及其关系的图形描述。传统知识图谱主要关注静态的常识性知识,缺乏对时间信息的考虑,难以处理网络空间中的动态演化和时间信息。时序知识图谱强调将时间纳入知识,通过引入时间戳和四元数嵌入管理动态时态知识,为紧密时间耦合的应用提供支持。现有方法侧重于高频重复事件,可能导致对新事件的错误判断。对此,我们提出了时空知识感知网络(SKAN),分为主体空间感知、事件序列感知和关系表征学习三个模块。SKAN通过全局关联图卷积挖掘主体关联性,进行时序学习,并通过关系表征学习预测未来关系,为时序知识图谱学习提供了新的架构。我们在四个国际事件数据集上进行了实验,实验结果表明,我们的方法优于目前的主流方法。
Knowledge graphs provide graphical representations of entities, concepts, and their relationships in the objective world. Traditional knowledge graphs mainly focus on static commonsense knowledge, lacking consideration for temporal information, making it challenging to handle dynamic evolution and temporal data in cyberspace. Temporal knowledge graphs emphasize incor-porating time into knowledge by introducing timestamps and quaternion embeddings to manage dynamic temporal knowledge, offering support for tightly time-coupled applications. Existing methods often prioritize high-frequency repetitive events, potentially leading to erroneous judgments for new events. To address this, we propose the Spatial-temporal Knowledge Aware Network (SKAN), divided into three modules: Subject Spatial Perception, Event Sequence Perception, and Relationship Representation Learning. SKAN utilizes global association graph convolution for subject correlation mining, conducts temporal learning, and predicts future relationships through relationship representation learning, providing a novel architecture for temporal knowledge graph learning. Experimental results on four international event datasets demonstrate the superiority of our approach over current mainstream methods.
[1] | 肖蕾, 李琪. 时序知识图谱补全方法研究综述[J/OL]. 计算机工程与应用: 1-14.
http://kns.cnki.net/kcms/detail/11.2127.TP.20230915.1820.012.html, 2024-01-23. |
[2] | Jiang, T., Liu, T., Ge, T., et al. (2019) Towards Time-Aware Knowledge Graph Completion. Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, Osaka, December 2016, 1715-1724. |
[3] | Zhu, C., Chen, M., Fan, C., et al. (2021) Learning from History: Modeling Temporal Knowledge Graphs with Sequential Copy-Generation Networks. Proceedings of the AAAI Conference on Artificial Intelligence, 35, 4732-4740.
https://doi.org/10.1609/aaai.v35i5.16604 |
[4] | Li, Z., Jin, X., Li, W., et al. (2021) Temporal Knowledge Graph Reasoning Based on Evolutional Representation Learning. Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval, New York, 11-15 July 2021, 408-417. https://doi.org/10.1145/3404835.3462963 |
[5] | Li, Z., Guan, S., Jin, X., et al. (2022) Complex Evolutional Pattern Learning for Temporal Knowledge Graph Reasoning. arXiv: 2203.07782. |