Exploiting second-order dissimilarity representations for hierarchical clustering and visualization - Data Mining and Knowledge Discovery