Semi-supervised Formality Style Transfer using Language Model Discriminator and Mutual Information Maximization

TL;DR

This paper introduces a semi-supervised style transfer model that uses a language model discriminator and mutual information maximization to convert informal sentences into formal ones, outperforming previous methods.

Contribution

It proposes a novel semi-supervised approach with a language model discriminator and mutual information maximization for style transfer, extending to unsupervised sentiment transfer.

Findings

Outperformed state-of-the-art baselines in formality transfer.

Achieved significant improvements on sentiment style transfer datasets.

Demonstrated effectiveness of mutual information maximization.

Abstract

Formality style transfer is the task of converting informal sentences to grammatically-correct formal sentences, which can be used to improve performance of many downstream NLP tasks. In this work, we propose a semi-supervised formality style transfer model that utilizes a language model-based discriminator to maximize the likelihood of the output sentence being formal, which allows us to use maximization of token-level conditional probabilities for training. We further propose to maximize mutual information between source and target styles as our training objective instead of maximizing the regular likelihood that often leads to repetitive and trivial generated responses. Experiments showed that our model outperformed previous state-of-the-art baselines significantly in terms of both automated metrics and human judgement. We further generalized our model to unsupervised text style transfer task, and achieved significant improvements on two benchmark sentiment style transfer datasets.