Applying Key Concepts Extraction for Evaluating the Quality of Students’ Highlights on e-Book

Abstract

The quality of students’ highlights can be an indicator of their learning performance. While the most common approach to grade their highlights is by humans, human grading can be inconsistent, especially when the number of highlights are large or when graders have different background knowledge. In this research, we propose a model to automatically extract important concepts from class materials, analyze students’ highlights and find the correlation between highlight quality and students’ learning performance. We first compared different text summarization algorithms with different evaluations to see which of them generates the summarization that is closest to the reference answer generated by humans. Then we used the selected algorithm to summarize the text from learning materials as important concepts, and compared the summaries with students’ highlights to calculate their highlight scores. Finally, we considered the highlight score from the best method as the highlight quality and observed whether it has a correlation to students’ learning performance.