ST$^2$: Small-data Text Style Transfer via Multi-task Meta-Learning
This addresses the challenge of few-shot style transfer for NLP applications, enabling more versatile text adaptation beyond limited categories like positive/negative.
The paper tackled the problem of text style transfer with limited data by developing a meta-learning framework that enables transfer between any kind of text styles, including fine-grained personal styles, resulting in improved language fluency and style transfer accuracy where state-of-the-art models fail.
Text style transfer aims to paraphrase a sentence in one style into another style while preserving content. Due to lack of parallel training data, state-of-art methods are unsupervised and rely on large datasets that share content. Furthermore, existing methods have been applied on very limited categories of styles such as positive/negative and formal/informal. In this work, we develop a meta-learning framework to transfer between any kind of text styles, including personal writing styles that are more fine-grained, share less content and have much smaller training data. While state-of-art models fail in the few-shot style transfer task, our framework effectively utilizes information from other styles to improve both language fluency and style transfer accuracy.