Towards automated complex ontology alignment using rule-based machine learning

Abstract

An ontology usually serves as the schema of a knowledge graph, which provides a vocabulary describing one or many domains of discourse and a specification of the meaning of terms in that vocabulary. Different parties would in general adopt different ontologies. And each ontology may have its own data vocabulary, modeling philosophy, and even language, which makes the semantic data integration process very challenging. To facilitate interoperability between different organizations, ontology alignment has been considered as the silver bullet for many applications. Ontology alignment has been studied for over a decade, and over that time many alignment systems and methods have been developed by researchers in order to find simple 1-to-1 equivalence matches between two ontologies. However, the simple correspondences are not expressive enough to fully cover the different types of heterogeneities in real-world problems. And very few alignment systems focus on finding complex correspondences. There are several reasons for this limitation. First, there are no widely accepted alignment benchmarks that contain such complex relationships. Second, tackling complex alignment is more challenging than finding simple alignment. It also requires experts from different domains to work together to manually generate the alignment, which is extremely time-consuming and inefficient. Third, the traditional evaluation metrics like precision, recall, and f-measure, are not fine-grained enough to evaluate the performance of complex alignment systems. Therefore, it hinders the generation and evaluation of complex ontology alignment systems. To tackle this problem and advance the development of ontology matching and alignment, we seek to address the problem by first developing potential benchmarks that contain the complex relations from real-world ontologies. We then propose an automated complex ontology alignment system based on association rule learning to generate not only simple correspondences but also complex ones. The algorithm can also be used in a semi-automated fashion to effectively assist users in finding potential complex alignments that they can then validate or edit. Finally, we evaluate the performance of the proposed algorithm on the benchmarks and analyze the results in detail and provide insights into the field of complex ontology alignment.