Students are one of the users of services provided by institutions. Student complaint questionnaires were distributed to find out various complaints related to institutional services. The problem is how to interpret student complaints so that institutional services can be improved according to what is needed. Therefore, this research aims to analyze the sentiment of student complaints. Apart form that, this research also aims to find the context of each student's complaint regarding institutional services. This research uses sentiment analysis with Indonesian text. The method starts from data collection, data cleaning and labelling of complaints, pre-processing of complaint text, term frequency (TF) method used to extract the content, and accuracy measurement. The labeling process was carried out twice to compare the accuracy, precision, recall, and f1-score of the models. Based on the results of the sentiment analysis of student complaints, the accuracy rate reached 76.1%. Additionally, a precision of 65.2% and recall of 85.2% indicate that labeling is more balanced, although there is still room for improvement.