On Learning Text Style Transfer with Direct Rewards

Abstract

In most cases, the lack of parallel corpora makes it impossible to directly train supervised models for the text style transfer task. In this paper, we explore training algorithms that instead optimize reward functions that explicitly consider different aspects of the style-transferred outputs. In particular, we leverage semantic similarity metrics originally used for fine-tuning neural machine translation models to explicitly assess the preservation of content between system outputs and input texts. We also investigate the potential weaknesses of the existing automatic metrics and propose efficient strategies of using these metrics for training. The experimental results show that our model provides significant gains in both automatic and human evaluation over strong baselines, indicating the effectiveness of our proposed methods and training strategies.

Published as a long paper at NAACL 2021

References in corpus (7)

• BERTScore: Evaluating Text Generation with BERT
• Unsupervised Learning of Sentence Embeddings using Compositional n-Gram Features
• Style Transfer from Non-Parallel Text by Cross-Alignment
• TransferTransfo: A Transfer Learning Approach for Neural Network Based Conversational Agents
• Relevance of Unsupervised Metrics in Task-Oriented Dialogue for Evaluating Natural Language Generation
• Style Transfer as Unsupervised Machine Translation
• A Probabilistic Formulation of Unsupervised Text Style Transfer