Formality Style Transfer with Hybrid Textual Annotations

TL;DR

This paper introduces a hybrid model for formality style transfer that leverages both parallel and classified data, achieving state-of-the-art results and adaptability to other style transfer tasks.

Contribution

The proposed omnivorous model effectively combines different data types to improve formality transfer and can be adapted to other style transfer tasks.

Findings

Achieved state-of-the-art performance on formality transfer benchmark.

Successfully adapted to sentiment transfer with competitive results.

Demonstrated effectiveness of hybrid data approach in style transfer.

Abstract

Formality style transformation is the task of modifying the formality of a given sentence without changing its content. Its challenge is the lack of large-scale sentence-aligned parallel data. In this paper, we propose an omnivorous model that takes parallel data and formality-classified data jointly to alleviate the data sparsity issue. We empirically demonstrate the effectiveness of our approach by achieving the state-of-art performance on a recently proposed benchmark dataset of formality transfer. Furthermore, our model can be readily adapted to other unsupervised text style transfer tasks like unsupervised sentiment transfer and achieve competitive results on three widely recognized benchmarks.