Abstract
The proliferation of video-sharing platforms and MOOCs has raised new challenges in the field of education. A challenging topic that is gaining an increased popularity is the identification of prerequisite relations between concepts in video lectures.
In this paper, we propose unsupervised methods for prerequisite identification and the creation of a prerequisite graph. The contribution, compared to existing approaches, is the development of methods which (i) do not rely on external knowledge, (ii) do not require extensive training, and (iii) are intended to exploit both the lecture transcript and its visual features.
Results from the preliminary evaluation are encouraging, and provide insights on the extraction of prerequisite relations from video transcripts compared to textbooks.