Building an Effective and Efficient Background Knowledge Resource to Enhance Ontology Matching

Abstract

Ontology matching is critical for data integration and interoperability. Original ontology matching approaches relied solely on the content of the ontologies to align. However, these approaches are less effective when equivalent concepts have dissimilar labels and are structured with different modeling views. To overcome this semantic heterogeneity, the community has turned to the use of external background knowledge resources. Several methods have been proposed to select ontologies, other than the ones to align, as background knowledge to enhance a given ontology-matching task. However, these methods return a set of complete ontologies, while, in most cases, only fragments of the returned ontologies are effective for discovering new mappings. In this article, we propose an approach to select and build a background knowledge resource with just the right concepts chosen from a set of ontologies, which improves efficiency without loss of effectiveness. The use of background knowledge in ontology matching is a double-edged sword: while it may increase recall (i.e., retrieve more correct mappings), it may lower precision (i.e., produce more incorrect mappings). Therefore, we propose two methods to select the most relevant mappings from the candidate ones: (1) a selection based on a set of rules and (2) a selection based on supervised machine learning. Our experiments, conducted on two Ontology Alignment Evaluation Initiative (OAEI) datasets, confirm the effectiveness and efficiency of our approach. Moreover, the F-measure values obtained with our approach are very competitive to those of the state-of-the-art matchers exploiting background knowledge resources.

Keywords: Ontology matching, Ontology alignment, Background knowledge, Indirect matching, External resource, Anchoring, Derivation, Background knowledge selection, Supervised machine learning