MIntRec2.0 is a large-scale multimodal multi-party benchmark dataset for intent recognition and out-of-scope detection in conversations.
We uploaded the data set MIntRec2.0 in two folders: data and data01 respectively.
Dataset Features
a.Large in Scale
Compared with our first version of multimodal intent recognition dataset (MIntRec), MIntRec2.0 increase the data-scale from 2.2K to 15K, with 30 intent classes, 9.3K in-scope and 5.7K out-of-scope annotated utterances with… See the full description on the dataset page: https://huggingface.co/datasets/THU-IAR/MIntRec2.0.
