MIntRec
收藏资源简介:
MIntRec数据集由清华大学智能技术与系统国家重点实验室开发,专注于多模态意图识别,包含2224个高质量样本,涵盖文本、视频和音频三种模态。数据集内容丰富,源自电视剧《超级商店》,通过精细的意图分类体系,包括2个粗粒度和20个细粒度意图类别,支持深入研究。创建过程中,研究团队采用了自动化的说话人标注流程,提高了标注效率和准确性。该数据集适用于提升意图识别的准确性和理解复杂人类意图的研究,特别是在多模态场景下的应用。
The MIntRec dataset, developed by the State Key Laboratory of Intelligent Technology and Systems at Tsinghua University, focuses on multimodal intent recognition. It includes 2,224 high-quality samples covering three modalities: text, video, and audio. Derived from the TV series *Superstore*, the dataset features a sophisticated intent classification taxonomy comprising 2 coarse-grained and 20 fine-grained intent categories, enabling in-depth research. During its development, the research team adopted an automated speaker annotation workflow to improve annotation efficiency and accuracy. This dataset is applicable to research aiming to enhance the accuracy of intent recognition and understand complex human intentions, especially for applications in multimodal scenarios.




