Topic-Level Social Network and Language Correlation in Course Discussion Forums

In this paper, we characterize student contributions in online discussion forums and examine the relationship between these characteristics and students' peer-to-peer relationships. We use Coh-Metrix Language and Discourse Analysis metrics to predict the Weighted Degree and Closeness Centrality of student contributions. We performed the analysis on all students in the population under study and on a subset consisting of active students only. We found that students who have more direct connections with other students tend to have abstract word choices whereas active students also tend to be more expository and informational, have shallow ideas, and use simpler construction. We also found that all students who can easily share their thoughts to the entire class tend to have posts that are more informational, with deep and connected thoughts and ideas. Active students who belong in this group further exhibit simpler construction and more abstract word choices.