STYLEPTB

STYLEPTB是一个大规模的文本风格转移基准数据集，由卡内基梅隆大学创建。该数据集包含59,767对句子，涉及21种细粒度的风格变化，包括词汇、句法、语义和主题层面的转移。数据集旨在通过提供细粒度的风格控制，推动可控文本生成和风格转移的研究。STYLEPTB不仅支持单一风格转移，还允许组合多种风格转移，为复杂、高层级的风格转移提供基础。数据集的应用领域广泛，旨在解决文本生成中的风格控制问题，提高生成文本的质量和多样性。

STYLEPTB is a large-scale text style transfer benchmark dataset developed by Carnegie Mellon University. This dataset contains 59,767 sentence pairs, involving 21 fine-grained style variations, with transfers covering lexical, syntactic, semantic, and thematic levels. The dataset is designed to advance research in controllable text generation and style transfer by providing fine-grained style control capabilities. STYLEPTB not only supports single-style transfer but also enables the combination of multiple style transfers, laying a solid foundation for complex, high-level style transfer tasks. It has a wide range of application scenarios, aiming to address style control challenges in text generation and improve the quality and diversity of generated texts.