OffVid: A System for Linking Off-Topic Concepts to Topically Relevant Video Lecture Segments

Abstract

We present a system for automatically connecting off-topic concepts from a video lecture to appropriate and topically relevant video lecture segments. The linked video lectures are expected to provide more detailed account of the corresponding off-topic concept. The system is realized with three modules: off-topic identification, topic base generation and segment linking. We modelled the problem of finding off-topic concept identification task as a community structure analysis in concept similarity graph. Word embedding-based technique has been used to generate topic specific video segments that act as the targets of off-topic concept connection candidates. The segmented videos are indexed using extracted topic by Solr search engine and are retrieved against queries formulated from offtopic concepts. The system has been evaluated on video lecture picked up from NPTEL MOOC platform. User study on the quality of recommendation has been found to be promising.