Visually exploring sentiment and keywords for analysing student satisfaction data

Measuring student satisfaction is an important task, which can lead to improvements in teaching strategies and techniques. There are several ways that satisfaction can be measured. One of the most common ways, is for students to rate or score a subject on a Likhert scale. This can give educators an initial grasp on subject satisfaction, however this doesn’t provide insight into reasons why the score was given. Free text comments are often given by students along with the score but summarising or finding meaningful information from these comments can be quite time consuming, especially for large classes. These comments can be used for educators to improve their teaching practice. In this study, we explored the use of machine learning techniques to visualise student satisfaction. This visualisation was exploited in the context of the following 2 research questions, 1. How can we visualise keywords, sentiment, and relationships between these keywords? 2. What benefit would this have for students and educators? Following on from previous work completed, we used several text analysis techniques to process initially clean and process text data. The dataset used for this analysis is student satisfaction survey data obtained from systematic university evaluations of units. These consist of both a satisfaction score, and a free text response, allowing students to give detailed feedback. We identified keywords, and the sentiment of these keywords, as well as position of these keywords to relate certain ideas together. This process is completed automatically, so a user or lecturer can use and benefit from the visualisations without understanding the technical aspects of this text analysis process. A high-level graph is generated, for each unit, which includes keywords, sentiment of the keywords, and demonstrates connections between the keywords. A large amount of comments were used, and analysed using the various methods presented. Results analysing units revealed that keywords such as “assessment” and “tutorials” were prominently featured in the generated visualisations. Using these generated visualisations, large amounts of comments are compared with various subjects. This comparison is valuable in identifying and linking key factors related to teaching and learning approaches used, as well as those relating to the teaching environment. These visualisations are showed in the context of engineering specific learning activities such as programming or electronic circuit analysis, and also compare the teaching activity terms between various engineering and non-engineering units. Pointers to certain approaches to teaching can have an impact for learning and allow the educator to ensure there is constructive alignment between learning activities, assessment and unit outcomes. This will improve students’ overall learning experience and outcomes.