ScholarWorks@숙명여자대학교

초록

Objectives This study explores how three text analysis techniques―TF-IDF, co-occurrence analysis, and topic modeling (LDA)―can effectively interpret unstructured student feedback collected during the “Check” phase of a PDCA cycle in institutional research (IR). It aims to enhance the validity of such feedback, reflect depart ment-specific needs, and inform improvements in the subsequent “Act” phase. Methods Student comments were gathered from three 2022 surveys at a four-year private university in Seoul: curriculum satisfaction, overall student satisfaction, and mid-term course evaluations. TF-IDF helped identify critical department-level issues, co-occurrence analysis uncovered interconnected themes, and LDA revealed students’ key concerns. These results guided administrative and departmental strategies for targeted improvements. Results First, TF-IDF illuminated each department’s unique priorities, enabling more precise problem identification. Second, co-occurrence analysis highlighted core topics requiring cross-departmental collaboration, clarifying in tegrated improvement tasks. Third, LDA-based topic modeling pinpointed immediate instructional strengths and weaknesses in mid-term evaluations, allowing prompt course adjustments and relevant support from admin istrative units. Conclusions Incorporating text analysis of unstructured student data significantly bolsters efforts to achieve stu dent-centered education. To maximize its impact, institutions should carefully plan data collection and pre processing, choose analysis methods suited to specific feedback contexts, and foster a culture of continuous im provement grounded in student perspectives. This approach can lead to meaningful insights for both policy deci sions and educational practice.

키워드