Structured Data for Content-Style Representation

该数据集包含结构化的数据，可进行分词处理，涵盖了文本、语音和图像等多种类型，旨在学习内容与风格分离的表示方法。此外，该数据集被用于评估所提出的不监督学习框架——检索器，在语音和图像领域的表现。该任务专注于内容与风格的分解。

This dataset contains structured data that supports tokenization, covering multiple modalities including text, speech and images. It is designed to learn disentangled representations of content and style. Furthermore, this dataset is utilized to evaluate the performance of the proposed unsupervised learning framework — Retriever — in the domains of speech and image. This task focuses on the decomposition of content and style.