Model Predicts Video-Induced Pleasure via Cognitive Appraisal

A novel computational model forecasts enjoyment derived from video content by examining cognitive appraisal factors. This model tackles four key issues: inconsistent human labeling, the disparity between positive emotions and pleasure, a lack of datasets focused on pleasure, and the opaque nature of black-box fusion techniques. It combines both data-driven and cognitive theory-based methods, employing cognitive appraisal theory alongside a fuzzy model within a transformer architecture that utilizes attention mechanisms for extracting multimodal features. This research was released on arXiv (ID: 2604.23753) as a recent announcement.