Clustering Based Recommendation of Pedagogical Resources

Abstract

In France, seven DTUs (Digital Thematic Universities) allow open access tomore than 24,000 OERs (Open Educational Resources). A DTU is a thematic repository of OERs, all validated by the academic community and indexed using SupLomFR (the French declaration for higher education of the LOM standard). The emergence of many huge repositories of OER offers new opportunities to learners, where the OER can be freely accessed from the DTU’s portal, at any moment, by anybody, from everywhere. But it is difficult for most of learners to find interesting resources, when the only available information about resources is their indexing in an international standard suchas the LOM or one of its national declarations. Thus it is important to helplearners to find pertinent resources, even if the only thing known about alearner is the last resource he selects in the current session. One way toperform an accurate recommendation is to recommend him the nearestresources to the last one he clicked on, in term of similarity. But the nearestresources can be highly dissimilar to the last clicked resource if this latter is anisolated one. In this case, it is better to recommend nothing to him, as we cannot afford to recommend an inappropriate material in an e-learningcontext.