On Variational Learning of Controllable Representations for Text without Supervision
This addresses the challenge of enabling controllable text manipulation without supervision, which is incremental but impactful for natural language processing applications.
The paper tackled the problem of unsupervised learning of controllable representations for text, finding that sequence VAEs fail due to latent vacancy issues, and proposed constraining the posterior mean to a learned probability simplex, achieving first success in this area with empirical outperformance on text style transfer and fine-grained control.
The variational autoencoder (VAE) can learn the manifold of natural images on certain datasets, as evidenced by meaningful interpolating or extrapolating in the continuous latent space. However, on discrete data such as text, it is unclear if unsupervised learning can discover similar latent space that allows controllable manipulation. In this work, we find that sequence VAEs trained on text fail to properly decode when the latent codes are manipulated, because the modified codes often land in holes or vacant regions in the aggregated posterior latent space, where the decoding network fails to generalize. Both as a validation of the explanation and as a fix to the problem, we propose to constrain the posterior mean to a learned probability simplex, and performs manipulation within this simplex. Our proposed method mitigates the latent vacancy problem and achieves the first success in unsupervised learning of controllable representations for text. Empirically, our method outperforms unsupervised baselines and strong supervised approaches on text style transfer, and is capable of performing more flexible fine-grained control over text generation than existing methods.