MOOC Discourse Analysis
Researchers from the CSER group have developed a visualisation dashboard to demonstrate the emergent themes from Massive Open Online Courses (MOOC) discussion contents.
Supported by a Google Research Award to explore personalised learning within MOOC environments, this work supports the automated clustering and labelling of discussion threads according to topics, which can then be used by learners to help them navigate the extensive resources and discussion content typically available in large scale MOOC environments. The dynamic and self-paced nature of MOOCs makes this an even more pressing need, in that learners follow their own pace through many MOOC courses, meaning that they may ask questions or seek advice on any course topic at any time. This work represents a first step towards assisting learners in finding personalised learning cohorts at scale.
In our initial results, published at Learning@Scale, we analysed three popular MOOCs from the Coursera platform – Machine Learning, Statistics, and Psychology to uncover latent discussion topics. The main contribution of this work involves proposing a topic labelling mechanism to label topic clusters generated from topic models (e.g. Latent Dirichlet Allocation) which is otherwise challenging due to the dynamic and diverse nature of discussions [1-2]. Our work introduces a topic visualisation dashboard which demonstrates the emergent topics and their relationship with discussions. The dashboard also capable of visualising the relationship between discussion topics and different variables associated with online discourse including views, votes, posts, instructor interventions, and time series analysis .
The screenshots below demonstrates the visualisation dashboard and sample topic maps.