Thank you BART! Rewarding Pre-Trained Models Improves Formality Style Transfer
This work addresses content preservation in formality style transfer, which is important for natural language processing applications, but it is incremental as it builds on existing pre-trained models.
The paper tackled the problem of content preservation in formality style transfer by fine-tuning pre-trained models (GPT-2 and BART) with style and content rewards, achieving a new state-of-the-art result.
Scarcity of parallel data causes formality style transfer models to have scarce success in preserving content. We show that fine-tuning pre-trained language (GPT-2) and sequence-to-sequence (BART) models boosts content preservation, and that this is possible even with limited amounts of parallel data. Augmenting these models with rewards that target style and content -- the two core aspects of the task -- we achieve a new state-of-the-art.