ERATO

ERATO数据集是由阿里巴巴集团创建的大型多模态数据集，专注于戏剧中人物间的情感关系识别。该数据集包含31,182个视频片段，总时长约203小时，涵盖多种模态信息，如视觉、音频和文本。数据集通过严格的质控流程收集和标注，确保高质量。ERATO数据集不仅推动了情感关系识别的研究，还促进了多模态融合方法的发展，特别是在视频理解领域中，为理解和分析人物关系提供了重要资源。

The ERATO dataset is a large-scale multimodal dataset created by Alibaba Group, focusing on emotional relationship recognition between characters in dramas. It contains 31,182 video clips with a total duration of approximately 203 hours, covering diverse modalities including visual, audio, and text. The dataset was collected and annotated through a strict quality control process to ensure high quality. The ERATO dataset not only advances research on emotional relationship recognition but also promotes the development of multimodal fusion methods, especially in the field of video understanding, providing an important resource for understanding and analyzing interpersonal relationships of characters.