mTVR

该数据集是一个大规模的多语言视频时刻检索数据集，包含了来自21.8K个电视节目视频片段的218K个英文和中文查询。该数据集在流行的TVR数据集基础上进行了扩展，增加了配对的中文查询和字幕。此外，数据集中包含了经过人工检查的英文查询的中文翻译，通过严格的标注过程确保了查询与视频内容的对齐和质量。该数据集的规模为218K个查询和21.8K个视频片段，其任务是视频时刻检索。

This dataset is a large-scale multilingual video moment retrieval dataset containing 218K English and Chinese queries derived from 21.8K TV program video clips. It is expanded from the widely-used TVR dataset, with supplementary paired Chinese queries and subtitles incorporated. Additionally, the dataset features human-validated Chinese translations of English queries, with a rigorous annotation process implemented to guarantee both the alignment between each query and its corresponding video content and the overall data quality. Boasting a scale of 218K queries and 21.8K video clips, this dataset is designed for the video moment retrieval task.