| Abstract |
Combining the output of several alignment tools is a crucial problem for ontology alignment technology. Extending on work that proposed preference-aware argumentation frameworks to deal with this problem, we have run a number of experiments with data from the National Libary of the Netherlands. We have especially investigated ways to consider quantitative aspects of aligner's outputs, either intrinsic (confidence levels) or extrinsic (consensus among mappers). I will introduce the approach we have turned to, present some results, and point at issues that arise when implementing the basics of an argumentation framework, for instance regarding the criteria to define whether an argument attacks another. |