An automatic question generation tool for supporting sourcing and integration in students' essays

Abstract

This paper presents a domain independent Automatic Question Generation (AQG) tool that generates questions which can be used as a form of support for students to revise their essay. The focus here is on generating questions based on semantic and syntactic information acquired from citations. The semantic information includes the author's name, the citation type (describing the aim of the cited study, its results or an opinion), the author's expressed sentiment, and the syntactic information of the citation. Pedagogically, the question templates are designed using Bloom's learning taxonomy where the questions reach the Analysis Level. We used 40 undergraduate students essays for our experiment and the Name Entity Recognition component is trained on 20 essays. The result of our experiment shows that the question coverage is 96% and accuracy of generated questions can reach 78%. This AQG tool will be integrated into our peer review system to scaffold feedback from peers.