Dimensional clustering of Linked Data : techniques and applications

The plurality and heterogeneity of linked data features require appropriate solutions for accurate matching and clustering. In this paper, we propose a dimensional clustering approach to enforce (i) the capability to select the set of features to use for data matching and clustering, that are packaged into the so-called thematic dimension, and (ii) the capability to make explicit the cause of similarity that generates each cluster. Ensemble techniques for combining different single-dimension cluster sets into a sort of multi-dimensional view of the considered linked data are also presented as a further contribution of the paper. Application to linked data summarization and exploration is finally discussed.