User Tagging in MOOCs Through Network Embedding

Tags are used everywhere nowadays in popular systems and play very important role due to the semantic meanings they provide. For example, people use hash tags in Twitter to group relevant topics together. XuetangX, which is one of the largest MOOC platforms in China, also assign tags to the available courses to illustrate their disciplinary affiliations. Data analysis shows that users tend to enroll in courses under the same tag group as their previous enrollments. In this work, we research the problem of user tagging employing the user-course-tag heterogeneous network from XuetangX, where the tags are course categories. We employ multiple network embedding models to learn vector representations of users, courses and tags, from the network user-course-tag, and find that representations learned by node2vec performs best. Offline experiments show that the resulting personalized tag recommendation can further help other tasks like course recommendation. This function has also been deployed online to XuetangX, and promising results have been observed.