Tracking the Dynamics of SPOC Discussion Forums: A Temporal Emotion-Topic Modeling Approach

Abstract

Small private online courses (SPOCs) have drawn widespread attention due to their adaptability to blended learning in formal education. As a type of interactive tool, many SPOC forums have stored rich textual data including focused learning content and feedback. However, currently, this information is used mainly for measuring students' activity levels instead of academic emotion and their evolutionary trends throughout a teaching period. This paper presents an unsupervised forum understanding model, namely the temporal emotion-topic model (TETM), to model time jointly with emotions and topics. This model is applied to track the evolutionary trends of pairs of {emotion, topic}, i.e., emot-topics. Especially, for each emotion, the most significant topic can be extracted and tracked across a semester. Finally, we investigate the emot-topics differences of different achievement levels of students, and examine the dynamics of emot-topics over course weeks to gain some insights about the relationship between time, emotion and learning achievement. The tracking of emot-topics could be valuable if users are informed of what students are concerned about and how these topics evolve in courses.