Measuring Conceptual Similarity in Ontologies: How Bad is a Cheap Measure?

ABSTRACT. Several attempts have been made to develop similarity measures for ontologies. Motivated by finding problems in existing measures, we design a new family of measures to address these problems. We carry out two sets of empirical studies. Firstly, to explore how good the new measures are by comparing them to human judgments of similarity. Secondly, we investigate how likely it is to encounter specific task-oriented problems when using a bad similarity measure.