Automatic Question Generation based on MOOC Video Subtitles and Knowledge Graph

With the popularity of MOOCs (Massive Open Online Courses), videos have gradually replaced textbooks as the most common educational resources. Quiz is an important teaching tool in online learning process [1], and manually constructing quiz questions is time-consuming and labor-intensive, so there is a pressing need for automatic question generation from video subtitles. The challenge of this problem is that video subtitles are different from textbooks considering their characteristics such as redundancy, colloquialism, and no sentence division. In this paper, we firstly use external wiki knowledge graph (WIKIDATA [2]) to extract facts of interest from video subtitles, and then propose a novel template-based method to generate quiz questions from knowledge graph. Compared to traditional template-based method [4], we make improvements on the accuracy and comprehensibility of the question. Final experiment on SimpleQuestions dataset [3] shows that the proposed method outperforms other competitive methods by +6% in terms of BLEU on the task of question generations from knowledge graph [4].