Graph kernels for task 1 and 2 of the linked data data-mining challenge 2013
Gerben Klaas Dirk de Vries · UvA-DARE (University of Amsterdam) · 2013
In this paper we present the application of two RDF graph kernels to task 1 and 2 of the linked data data-mining challenge. Both graph kernels use term vectors to handle RDF literals. Based on experiments with the task data, we use the Weisfeiler-Lehman RDF graph kernel for task 1 and the intersection path tree kernel for task 2 in our final classiers for the challenge. Applying these graph kernels is very straightforward and requires (almost) no preprocessing of the data.