Text-mining Analyses of Undergraduates’ Essays on Global Perspectives by Performance Levels in Korea

Abstract

This study aims to illustrate how text-mining can be employed to identify key qualities of student performance on essays according to different performance levels. A total of 111 undergraduates’ essays on ‘climate change and transnational cooperation’ were classified into the upper, middle, and lower levels based on scores. The main findings from keyword frequency and network analyses are as follows. First, the contents of frequently used words were different across the performance levels. In the upper-level answers, the frequency of keywords used to analyze conflicts of interest between countries and to suggest cross-border responses using technology was high. In contrast, in the lower-level answers, everyday words often appeared to describe the need to respond to climate crisis. Second, in the upper-level answers, the weighted degree of centrality was similar across keywords, and the connections between keywords were stronger. While in the lower-level answers, the opposites were observed. Third, bigram network analysis was effective at all levels in identifying the structure of detailed discussions of essays. The usefulness and future directions of using text mining to analyze the qualitative difference in the subjects’ performance in essay assessment were discussed.