Multi-modal Synergy: Bridging Chinese Culture Expression and Teaching Interaction in Art English Textbooks via Self-supervised Learning

This dataset is optimized for English textbooks in art colleges, covering three modes of text, image and video. It contains 300,000 texts, 80,000 images and 2,000 hours of video, and the content involves multi-domain knowledge of art textbooks. In addition, 10,000 texts, 5000 images and 500 hours of videos were added from social media to test the performance of the model on noisy data, aiming to improve the semantic consistency, cultural expression integrity and teaching interactivity of textbooks, and provide strong support for the research on cross-cultural communication and multimodal integration of art English textbooks.

本数据集专为艺术院校的英语教材定制优化，涵盖文本、图像、视频三种模态。数据集包含30万条文本、8万幅图像以及2000小时的视频内容，覆盖艺术教材的多领域知识范畴。此外，额外从社交媒体补充采集了1万条文本、5000幅图像与500小时视频，用于测试模型在噪声数据下的性能表现，旨在提升教材的语义一致性、文化表达完整性与教学交互性，为艺术英语教材的跨文化传播与多模态融合研究提供坚实支撑。