Using Semi-supervised Deep Learning for Identifying Cognitive Engagement in Online Learning Discussion

Interactive discussion is an important component of online learning. Interactive discourses reflect different levels of cognitive engagement, and analyzing learners' cognitive engagement levels in discourses is beneficial for teachers to grasp learners' learning status and improve the course quality. However, using traditional automatic text classification models to analyze cognitive engagement requires a large amount of labeled data, which is time-consuming and laborious. In this paper, we propose a semi-supervised training method based on Linguistic Inquiry and Word Count (LIWC) and Unsupervised Data Augmentation (UDA) strategy for the cognitive engagement classification task, which aims to identify the cognitive engagement level of interactive discourse using a small amount of labeled data. The result shows that the model trained by the proposed method outperforms the benchmark models. Finally, the identification result is visualized to help to understand cognitive engagement structure and dynamics in different activity levels of learners.