Polarized-VAE: Proximity Based Disentangled Representation Learning for Text Generation
This work addresses the challenge of disentangling attributes in text data for improved generation, offering a general framework applicable to various tasks, though it appears incremental compared to existing unsupervised methods.
The authors tackled the problem of learning disentangled representations for text generation by proposing polarized-VAE, which uses proximity measures to separate attributes like semantics and syntax in the latent space. The method outperformed VAE baselines and was competitive with state-of-the-art approaches in transfer experiments.
Learning disentangled representations of real-world data is a challenging open problem. Most previous methods have focused on either supervised approaches which use attribute labels or unsupervised approaches that manipulate the factorization in the latent space of models such as the variational autoencoder (VAE) by training with task-specific losses. In this work, we propose polarized-VAE, an approach that disentangles select attributes in the latent space based on proximity measures reflecting the similarity between data points with respect to these attributes. We apply our method to disentangle the semantics and syntax of sentences and carry out transfer experiments. Polarized-VAE outperforms the VAE baseline and is competitive with state-of-the-art approaches, while being more a general framework that is applicable to other attribute disentanglement tasks.