Reducing the Risk of Completeness Loss by Subgraph Reasoning in Engineering Applications

Abstract

The ontology knowledge base is divided into TBox and ABox, the former is the schema-level information of the knowledge base, which is used to describe the relationship between the recognized concepts and attributes, and the latter is a collection of instance assertions or fact statements in the domain. Using TBox, the process of reasoning more implicit assertions in ABox is called ABox materialization, which plays an important role in knowledge base applications. Some ontology reasoners for OWL DL and DL-safe SWRL, the reasoning results are considered to be correct and complete. However, in practical applications, this paper finds that in some cases, the reasoning results are not complete when reasoning about the ontology knowledge base constructed by OWL DL and DL-sale SWRL. In those application scenarios that require completeness, blindly trusting these reasoners will bring risks. In response to this problem, this paper proposes a method based on subgraph reasoning, which can identify whether the result of ontology reasoning is complete in engineering, and at the same time, when it is incomplete, new assertions that have not been found before can be inferred. The comparative experimental results on two open-source ontologies verify the method in this paper.