Semi-supervised Formality Style Transfer using Language Model Discriminator and Mutual Information Maximization
This addresses the problem of generating formal sentences for NLP applications, with incremental improvements in style transfer tasks.
The paper tackled formality style transfer by proposing a semi-supervised model using a language model discriminator and mutual information maximization, which outperformed previous state-of-the-art baselines significantly in automated metrics and human judgment.
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.