The higher educational institutions gather student feedback after the end of each semester to improve the quality of education. The feedback consists of a grading scale to answer the questions followed by a textual response conveying the sentiments regarding the student’s experience. Since there is a considerable amount of response influx, going through every single textual feedback is time consuming; hence the need arises to extract sentiments from individual comments and classify them as positive, negative or neutral. The aim of our research is the comparison of various machine learning and deep learning approaches for developing an effective sentiment classification system for instructors. In this study, we analyzed student feedback consisting of 19000 comments and trained various machine learning and deep learning algorithms using several feature extraction techniques. Among the different algorithms employed, a cascading neural network consisting of CNN combined with LSTM using Glove word embedding outperformed all the other architectures giving an accuracy of 91.27%.  

Index Terms- Deep learning, feedback analysis, opinion mining, sentiment analysis, text mining