CMU-MOSEI

The dataset contains 250 different topics and 23,453 carefully labeled video clips with sentiment labels. The sentiment labeling of these clips follows the same criteria as CMU-MOSI, i.e., each discourse clip is assigned a sentiment value label ranging from -3 to +3, with a total of 7 levels.

Among them, -3 represents strong negative emotions, +3 points to strong positive emotions, and the gradual transition between the intermediate levels demonstrates the delicacy and comprehensiveness of emotional expression. The construction of this dataset provides richer and more diversified material resources for the research in the field of emotion calculation and understanding.