2017 Volume 41 Issue 3 Pages 233-244
This study aimed to analyze nine years’ worth of sixty thousand free descriptions of course evaluation questionnaires, which originally appeared to be overly difficult to handle in terms of extracting information that had been administered to a certain University Students using a topic model based on Latent Dirichlet Allocation (LDA). We extracted 170 topics and labeled them based on a series of steps, and then checked the validity. Consequently, an adequate level of validity was observed, revealing a comprehensible classification by topic model that fit the human senses. This study further offered analyses linking information about the group of courses to which each free description belonged, and visualized the ratio of topics in each course group, or features of topic distribution of each course group compared to the entire spectrum using cross tabulation. Such analyses are expected to be applied in the fields of Institutional Research (IR) and Learning Analytics (LA) in future studies.
