Disentangled Latent Representation Learning

AbstractThis paper tackles the problem of disentangling the latent representations of style and content in language models. We propose a simple yet effective approach, which incorporates auxiliary multi-task and adversarial …...