CLEFT: Contextualised Unified Learning of User Engagement in Video Lectures With Feedback

This paper addresses the challenge of predicting contextualized engagement in online teaching videos and providing constructive feedback to content creators. The proposed unified model, CLEFT (Contextualised unified Learning of user Engagement in video lectures with Feedback), utilizes multi-modal features, including language complexity, context information, textual emotion, animation, speaker’s pitch, and speech emotions. The ensemble of classifiers reliably detects engagement and offers valuable insights for content improvement.