Leveraging Book Indexes for Automatic Extraction of Concepts in MOOCs

Abstract

Concepts are basic elements in any learning module and are thus very useful for modeling, summarizing, and previewing the content of any module. Automatic extraction of the major concepts from online education materials enables many useful applications. In this paper, we propose to leverage textbooks and their back-of-the-book indexes as training data to train a supervised machine learning algorithm for automatic extraction of concepts from text data in the education domain. We evaluate this idea by training neural networks on three textbooks and applying the trained neural networks to extract concepts from the lecture transcripts of two MOOCs. Our results suggest great promise for further exploration of this direction.