Approximating Theoretical Linguistics Classification in Real Data: the Case of German “nach” Particle Verbs