Content preserving text generation with attribute controls

TL;DR

This paper presents a novel method for content-preserving text generation that allows for attribute control, ensuring generated sentences are fluent, attribute-compatible, and can handle multiple attributes simultaneously.

Contribution

The authors introduce a new model combining reconstruction and adversarial losses for improved attribute-controlled text generation, including multi-attribute control capabilities.

Findings

Generated sentences are more fluent and attribute-aligned.

The model outperforms prior methods in quality and control accuracy.

It successfully manages multiple attribute controls simultaneously.

Abstract

In this work, we address the problem of modifying textual attributes of sentences. Given an input sentence and a set of attribute labels, we attempt to generate sentences that are compatible with the conditioning information. To ensure that the model generates content compatible sentences, we introduce a reconstruction loss which interpolates between auto-encoding and back-translation loss components. We propose an adversarial loss to enforce generated samples to be attribute compatible and realistic. Through quantitative, qualitative and human evaluations we demonstrate that our model is capable of generating fluent sentences that better reflect the conditioning information compared to prior methods. We further demonstrate that the model is capable of simultaneously controlling multiple attributes.