We propose to employ graph kernels for clustering software modeling artifacts. Among other benefits, this would