Using Quantitative Aspects of Alignment Generation for Argumentation on Mappings

Abstract

State-of-the art mappers articulate several techniques using different sources of knowledge in an unified process. An important issue of ontology mapping is to find ways of choosing among many techniques and their variations, and then combining their results. For this, an innovative and promising option is to use frameworks dealing with arguments for or against correspondences. In this paper, we re-use an argumentation framework that considers the confidence levels of mapping arguments. We also propose new frameworks that use voting as a way to cope with various degrees of consensus among arguments. We compare these frameworks by evaluating their application to a range of individual mappers, in the context of a real-world library case.