Bag-of-Vectors Autoencoders for Unsupervised Conditional Text Generation

Florian Mai, James Henderson

Keywords: autoencoders, latent space learning, variable-size, natural language processing

Abstract: Text autoencoders are often used for unsupervised conditional text generation by applying mappings in the latent space to change attributes to the desired values. Recently, Mai et al. (2020) proposed $\operatorname{Emb2Emb}$, a method to $\textit{learn}$ these mappings in the embedding space of an autoencoder. However, their method is restricted to autoencoders with a single-vector embedding, which limits how much information can be retained. We address this issue by extending their method to $\textit{Bag-of-Vectors Autoencoders}$ (BoV-AEs), which encode the text into a variable-size bag of vectors that grows with the size of the text, as in attention-based models. This allows to encode and reconstruct much longer texts than standard autoencoders. Analogous to conventional autoencoders, we propose regularization techniques that facilitate learning meaningful operations in the latent space. Finally, we adapt $\operatorname{Emb2Emb}$ for a training scheme that learns to map an input bag to an output bag, including a novel loss function and neural architecture. Our experimental evaluations on unsupervised sentiment transfer and sentence summarization show that our method performs substantially better than a standard autoencoder.

One-sentence Summary: We represent text as a set of vectors and present a method to learn unsupervised sequence-to-sequence tasks as a mapping from one set of vectors to another set of vectors.

Community Implementations: [ 2 code implementations](https://www.catalyzex.com/paper/arxiv:2110.07002/code)