Knowledge is the cornerstone of artificial intelligence, which is often represented as RDF graphs. The large-scale RDF graphs in various fields pose new challenges to graph data management. Due to the maturity and stability, relational database is a good choice for RDF graph storage. However, the management of the complex structure of RDF graphs in the relational database requires sophisticated storage structure design. To address this problem, this paper adopts reinforcement learning (RL) to optimize the storage partition method of RDF graph. To the best of our knowledge, this is the first work to adopt RL to solve this problem. Moreover, we propose the featurization method of RDF tables which guarantees adequacy of state representation and the query rewriting policy which ensures correct query results when the storage structure changes. Extensive experiments on various RDF benchmarks demonstrate that the proposed approach significantly outperforms the state-of-the-art storage strategies.

知识是人工智能的基石，通常用RDF图来表示。各个领域的大规模 RDF 图对图数据管理提出了新的挑战。由于成熟度和稳定性，关系型数据库是 RDF 图存储的不错选择。然而，关系数据库中RDF图的复杂结构的管理需要复杂的存储结构设计。针对这个问题，本文采用强化学习（RL）来优化RDF图的存储分区方法。据我们所知，这是第一个采用 RL 来解决这个问题的工作。此外，我们提出了 RDF 表的特征化方法来保证状态表示的充分性，以及在存储结构发生变化时确保正确查询结果的查询重写策略。