ch-sims

The CH-SIMS dataset is a Chinese public dataset that includes a multimodal emotional corpus, consisting of 2281 video clips and covering 60 original videos. The dataset provides rich background information and a wide range of ages, as well as multimodal and tri modal single peak annotations. Each video clip is labeled with emotional labels, including regression labels and classification labels. Among them, the values of the regression label are {-1.0, -0.8, -0.6, -0.4, -0.2, 0.0, 0.2, 0.4, 0.6, 0.8, 1.0}, indicating the degree of emotional change from negative to positive. The classification labels are divided into -1 (negative), 0 (neutral), and 1 (positive). These labels can be used to evaluate the performance of sentiment analysis algorithms.