Automated Feedback Generation for Student Project Reports: A Data-Driven Approach

Instant feedback plays a vital role in promoting academic achievement and student success. In practice,
however, delivering timely feedback to students can be challenging for instructors for a variety of reasons
(e.g., limited teaching resources). In many cases, feedback arrives too late for learners to act on
the advice and reinforce their understanding. To this end, researchers have designed various automated
feedback systems in different domains, including novice programming, short-essay writing, and openended
questions. To the best of our knowledge, no previous study has investigated automated feedback
generation for a more complex form of student work - student project reports. In this work, we present
a novel data-driven system, dubbed Insta-Reviewer, for automatically generating instant feedback on
student project reports. In addition to automatic metrics such as ROUGE scores and BERTScore, we
propose a five-dimension framework for manually evaluating system-generated feedback. Experimental
results show that feedback generated by Insta-Reviewer on real students’ project reports can achieve
near-human quality. Our work demonstrates the feasibility of automatic feedback generation for students’
project reports while highlighting several prominent challenges for future research.