Geller, Shay A., et al. "New methods for confusion detection in course forums: student, teacher, and machine." IEEE Transactions on Learning Technologies 14.5 (2021): 665-679.

This article presents both computational and rule-based methods for detecting confusion in students' comments on course forums. To create reliable ground truth data for this task, a decision tree was designed to help experts manually label forum posts, though this process is time-consuming. To address this, confusion-detection rules were generated based on student hashtags, which were then applied to a large dataset from a biology course to automatically identify "confused posts." The dataset was used to train a BERT-based machine learning model, which outperformed traditional models in detecting confusion and demonstrated the ability to generalize across different course offerings. This approach offers a promising way to enhance confusion detection and support in online educational forums.