Usage-Based Clustering of Learning Resources to Improve Recommendations

In this paper, we introduce a usage-based technique for clustering learning resources accessed in online learning portals. This approach solely relies on the usage of the learning resources and does not consider their content or the relations between the users and the resources. In order to cluster the resources, we calculate higher-order co-occurrences, a technique taken from corpus-driven lexicology where it is used to cluster words based on their usage in language. We first outline how we adapt the approach to then present an extensive evaluation that shows the effects of the clustering. Finally, we show how the resulting clusters can be used to enhance recommender systems.