Educating Text Autoencoders: Latent Representation Guidance via Denoising

TL;DR

This paper introduces a denoising objective for autoencoders to improve the structure of latent spaces, enabling more coherent text manipulation and zero-shot style transfer.

Contribution

It proposes a denoising adversarial autoencoder that guides latent space geometry, enhancing controllable text generation and style transfer capabilities.

Findings

Improved latent space structure for text autoencoders

Enhanced zero-shot style transfer via latent vector arithmetic

Better trade-off between generation quality and reconstruction

Abstract

Generative autoencoders offer a promising approach for controllable text generation by leveraging their latent sentence representations. However, current models struggle to maintain coherent latent spaces required to perform meaningful text manipulations via latent vector operations. Specifically, we demonstrate by example that neural encoders do not necessarily map similar sentences to nearby latent vectors. A theoretical explanation for this phenomenon establishes that high capacity autoencoders can learn an arbitrary mapping between sequences and associated latent representations. To remedy this issue, we augment adversarial autoencoders with a denoising objective where original sentences are reconstructed from perturbed versions (referred to as DAAE). We prove that this simple modification guides the latent space geometry of the resulting model by encouraging the encoder to map similar texts to similar latent representations. In empirical comparisons with various types of autoencoders, our model provides the best trade-off between generation quality and reconstruction capacity. Moreover, the improved geometry of the DAAE latent space enables zero-shot text style transfer via simple latent vector arithmetic.