Opinion corpus for assessment of study abroad program

Abstract

This study compiled an opinion corpus for developing a method for automatically evaluating a study-abroad program. Evaluation should cover not only academic experience at a host institution but also intercultural experience in the dormitory and interpersonal experience with local students, which helps improve a study-abroad program. The corpus included 600 students’ opinions on the satisfaction with academic, intercultural and interpersonal experiences, consisting of 40,024 words in total. Each opinion was annotated according to the opinion polarity determined by an existing sentiment classifier automatically. When automatically classified opinion polarity was compared with manually determined opinion polarity, a different distribution was observed. Because the existing classifier was not trained with a corpus that dealt with the issues related to students’ opinions about a study-abroad program, this result suggested the need of a corpus for study-abroad program evaluation. The opinion classifier of this study trained with the opinion corpus demonstrated a higher accuracy (83.5 percent) than the majority class baseline (70.9 percent).