Extended Tversky Similarity for Resolving Terminological Heterogeneities across Ontologies

摘要

We propose a novel method to compute similarity between cross-ontology concepts based on the amount of overlap of the information content of their labels. We extend Tversky's similarity measure by using the information content of each term within an ontology label both for the similarity computation and for the weight assignment to tokens. The approach is suitable for handling compound labels. Our experiments showed that it outperforms existing terminological similarity measures for the ontology matching task.