A domain-dependent lexicon to augment CAD peer review

Abstract

CAD course creative projects necessitate subjective feedback. In academia, peer review is a widely used instrument to gather diverse and timely feedback which stimulates learning and engagement in students who review one another. To date, however, no effort to summarize and score subjective content from peer review text via sentiment analysis has been attempted in an educational setting, including CAD courses-many of which naturally employ a project-based architecture. This is perhaps due in part to a lack of specifically tuned tools. Towards meeting this need, we introduce a new lexicon compiled from actual peer review text, implemented specifically in a CAD-course context, and compare it to other publicly available lexicons. HeLPS, our domain-dependent lexicon, performed more concisely and accurately in our CAD courses and consistently tagged high-quality positive and negative sentiment with a lexicon a fraction of the size of others. Both qualitative and quantitative evidence suggest that HeLPS is the preferred option for identifying subjective opinion towards CAD course projects.