"MEED: A Multimodal Conversational Dataset for Learning-Related Emotion Recognition in Collaborative Learning"

"MEED (Multimodal Epistemic Emotion Dataset) is a multimodal emotional dataset collected from authentic face-to-face Collaborative Problem Solving (CPS) learning environments.The dataset captures natural student interactions during group-based problem-solving tasks and provides temporally aligned multimodal features at the utterance level.Modalities Included\ud83d\udc41 Visual modality \u2013 video-derived behavioral features\ud83c\udf99 Acoustic modality \u2013 speech-based acoustic features\ud83d\udcac Text modality \u2013 utterance-level transcripts and text embeddingsAll modalities are strictly aligned along the temporal dimension and segmented at the dialogue-utterance level.Each utterance is annotated into one of five epistemic emotion categories.These categories are grounded in epistemic emotion theories and reflect students' emotional states during knowledge construction and problem-solving.The dataset supports:Unimodal emotion recognitionMultimodal fusion modelingMulti-turn dialogue-level emotion dynamics modelingCognitive-affective interaction research"