Evaluating a Confidence Value for Ontology Alignment

Many methods for automatic and semi-automatic ontology alignment have been proposed, but they remain error prone and laborintensive. This paper describes a novel generic process for evaluating the mappings’ confidence value. This process uses rules extracted through inductive machine learning methods from the matching results proposed by others. Further, the precision and recall of the extracted rules are exploited in order to transform each rule into a mathematical formula that generates the mappings’ confidence value. Mappings are then classified not as valid or invalid but through a quantitative confidence value that can be easily managed during the alignment process.