Natural Language Processing of Student's Feedback to Instructors: A Systematic Review

Course developers, providers and instructors gather feedback from students to gain insights into student satisfaction, success, and difficulties in the learning process. The traditional manual analysis is time-consuming and resource-intensive, resulting in decreased insights and pedagogical impact. To address the problems, researchers use natural language processing techniques that apply the fields of machine learning, statistics and artificial intelligence to the feedback datasets for various purposes. These purposes include predicting sentiment, opinion research, insights into students' views of the course, and so on. The aim of this study is to identify themes and categories in academic research reports that use natural language processing for student feedback. Previous review studies have focused exclusively on sentiment analysis and specific techniques, such as machine learning and deep learning. Our study put forward a comprehensive synthesis of various aspects, from the data to the methods used, to the data translation and labeling efforts, and to the categorization of prediction/analysis targets in the literature. The synthesis includes two tables that allow the reader to compare the studies themselves and present the identified themes and categorizations in one figure and text. The methods, tools and data of 28 peer-reviewed papers are synthesized in 20 categories under six themes: aim and categorization, methods and models, and tools and data (size and context, language, and labeling). Our research findings presented in this article can inform researchers in the field in structuring their research ideas and methods, and in identifying gaps and needs in the literature for further development.