Linguistic analysis for complex ontology matching
Dominique Ritze, Johanna Völker, Christian Meilicke, Ondřej Šváb-Zamazal · MADOC (University of Mannheim) · 2010
Current ontology matching techniques focus on detecting correspondences between atomic concepts and properties.Nevertheless, it is necessary and possible to detect correspondences between complex concept or property descriptions.In this paper, we demonstrate how complex matching can benefit from natural language processing techniques, and propose an enriched set of correspondence patterns leveraging linguistic matching conditions.After elaborating on the integration of methods for the linguistic analysis of textual labels with an existing framework for detecting complex correspondences, we present the results of an experimental evaluation on an OAEI dataset.The results of our experiments indicate a large increase of precision as compared to the original approach, which was based on similarity measures and thresholds.