Evaluation Measures for Ontology Matchers in Supervised Matching Scenarios

摘要

Precision and Recall, as well as their combination in terms of F-Measure, are widely used measures in computer science and generally applied to evaluate the overall performance of ontology matchers in fully automatic, unsu-pervised scenarios. In this paper, we investigate the case of supervised matching, where automatically created ontology alignments are verified by an expert. We motivate and describe this use case and its characteristics and discuss why traditional, F-measure based evaluation measures are not suitable for this use case. Therefore, we investigate several alternative evaluation measures and propose the use of Precision@N curves as a mean to assess different matching systems for supervised matching. We compare the ranking of several state of the art matchers using Precision@N curves to the traditional F-measure based ranking, and discuss means to combine matchers in a way that optimizes the user support in supervised ontology matching.