Injecting Background Knowledge in Neural Models via Adversarial Set Regularisation - View it on GitHub