Incremental Reasoning for Dynamic ABox Materialization Based on Subgraph
Xixi Zhu, Bin Liu, Zhaoyun Ding, Li Yao, Cheng Zhu, Xianqiang Zhu · 2022
The ontology knowledge base mainly includes: TBox and ABox, the former models schema-level knowledge in the domain, and the latter is a statement of a set of instance assertions or facts. ABox materialization is the process of discovering the implicit knowledge in ABox by reasoning based on the existing knowledge, which has important value in the application of knowledge base. In practice, TBox has stability and durability, and is suitable for one-time construction and repeated use, while ABox is usually in a state of continuous growth and change. Even with efficient reasoning techniques, recomputing all possible results whenever the ABox changes is a time-and-resource-intensive operation, which can cause a significant delay, and is unacceptable for some application scenarios. In response to this problem, this paper proposes a subgraph-based incremental reasoning method for dynamic ABox materialization, which first divides the ABox into instance-centric subgraphs, and uses the collection of extended subgraphs of each instance that reasoning under the complete TBox as the materialized result. When the assertion is changed, the impact on the materialization result can be regarded as a limited step of information propagation between subgraphs, and only a small number of relevant subgraphs need to be re-derived, thus realizing the incremental reasoning of ABox materialization. The experimental results on two open-source ontology knowledge bases show that the method in this paper has high reasoning quality, and can effectively reduce the response time for obtaining materialized results when ABox changes dynamically.