Multi-Pair Text Style Transfer on Unbalanced Data
This work addresses text style transfer for applications with multiple, unbalanced data sources, though it appears incremental as it builds on existing nonparallel data methods.
The paper tackles multi-pair text style transfer on unbalanced data by developing a task-adaptive meta-learning framework that uses a single model, resulting in better quantitative performance and coherent style variations while handling data imbalance and domain mismatches.
Text-style transfer aims to convert text given in one domain into another by paraphrasing the sentence or substituting the keywords without altering the content. By necessity, state-of-the-art methods have evolved to accommodate nonparallel training data, as it is frequently the case there are multiple data sources of unequal size, with a mixture of labeled and unlabeled sentences. Moreover, the inherent style defined within each source might be distinct. A generic bidirectional (e.g., formal $\Leftrightarrow$ informal) style transfer regardless of different groups may not generalize well to different applications. In this work, we developed a task adaptive meta-learning framework that can simultaneously perform a multi-pair text-style transfer using a single model. The proposed method can adaptively balance the difference of meta-knowledge across multiple tasks. Results show that our method leads to better quantitative performance as well as coherent style variations. Common challenges of unbalanced data and mismatched domains are handled well by this method.