DRAG: Director-Generator Language Modelling Framework for Non-Parallel Author Stylized Rewriting

TL;DR

This paper introduces DRAG, a novel framework for author stylized rewriting that effectively controls target attributes and performs well with limited data, improving style accuracy and content preservation.

Contribution

The paper proposes a Director-Generator framework that enhances author stylized rewriting by explicitly controlling attributes and working with small corpora.

Findings

Significant improvements over existing methods in style accuracy.

Better content preservation and fluency in generated texts.

Effective with limited-sized target author corpora.

Abstract

Author stylized rewriting is the task of rewriting an input text in a particular author's style. Recent works in this area have leveraged Transformer-based language models in a denoising autoencoder setup to generate author stylized text without relying on a parallel corpus of data. However, these approaches are limited by the lack of explicit control of target attributes and being entirely data-driven. In this paper, we propose a Director-Generator framework to rewrite content in the target author's style, specifically focusing on certain target attributes. We show that our proposed framework works well even with a limited-sized target author corpus. Our experiments on corpora consisting of relatively small-sized text authored by three distinct authors show significant improvements upon existing works to rewrite input texts in target author's style. Our quantitative and qualitative analyses further show that our model has better meaning retention and results in more fluent generations.