A novel method to evaluate students sentiments from twitter messages

Abstract

Classroom education is a dynamic environment, which brings students from different backgrounds with diverse abilities. By introducing machine-learning algorithms to learn the sentiments of the students in a classroom-based environment, can provide us a better research tool to understand the student psychology behind their attentiveness as well as the impact the instructor has on them while delivering lectures. Emotions can be analyzed mainly through many ways like facial features, audio signals and text messages. In this study, we have proposed a student emotion classifying mechanism that works after the lecture by analyzing the tweets posted by the students in the social media platform, Twitter to study their sentiments, or thoughts as expressed in the department twitter handle as a feedback to the classroom lecture. Students can post a tweet to their respective department’s twitter handle about their opinions, emotions, suggestion. Our application has been designed to monitor the department’s handle, a unique user-id via twitter API handler and when any posts appear, collects it and predicts the emotion. A hybrid-based approach which contains lexical and learning based approaches will be used to handle the twitter-based data and to predict the emotions of a student. A lexicon dictionary will be used in lexical based approach and for learning based approach, a manually customized dataset was used, and a support vector machine was designed to train the datasets and classify the emotions. The use-case of this application can be ideal for colleges, companies and wherever anyone wants to ease up the process of analyzing the feedback, suggestions or complaints from the students or employees, thereby saving considerable manpower and time. Our proposal is expected to garner good results and improved prediction time and accuracy.