Cycle Consistent Adversarial AutoEncoders

AbstractUnsupervised text style transfer is full of challenges due to the lack of parallel data and difficulties in content preservation. In this paper, we propose a novel neural approach to …...