A computational approach to analyzing students' discussions on social media platforms

Abstract

Although many studies have analyzed social media usage in various educational contexts, few have applied computational methods to scale up the analysis so that it could be adopted more widely in education practice. To bridge the gap, this study applies three computational methods, namely temporal mining of themes, sentiment analysis, and sequential pattern mining to compare student collaborative learning patterns on two popular social media platforms, Blog and Facebook. This study found that both Facebook and Blog supported student reflections and discussions on similar topics, but on Facebook negative postings appeared in an earlier stage and students had more interactions. This study showcases that computational methods are scalable and efficient and thus can help educators analyze and optimize students' learning in a timely manner.