Plug and Play Autoencoders for Conditional Text Generation
This work addresses the challenge of efficient and data-efficient conditional text generation for tasks like style transfer, though it is incremental as it builds on existing autoencoder techniques.
The paper tackles the problem of conditional text generation by proposing a plug-and-play method that uses any pretrained autoencoder and trains an embedding-to-embedding mapping, reducing the need for labeled data and improving efficiency. The method achieves performance comparable to strong baselines on style transfer tasks while being up to four times faster.
Text autoencoders are commonly used for conditional generation tasks such as style transfer. We propose methods which are plug and play, where any pretrained autoencoder can be used, and only require learning a mapping within the autoencoder's embedding space, training embedding-to-embedding (Emb2Emb). This reduces the need for labeled training data for the task and makes the training procedure more efficient. Crucial to the success of this method is a loss term for keeping the mapped embedding on the manifold of the autoencoder and a mapping which is trained to navigate the manifold by learning offset vectors. Evaluations on style transfer tasks both with and without sequence-to-sequence supervision show that our method performs better than or comparable to strong baselines while being up to four times faster.