Understanding students’ learning experiences by mining social media data

Abstract

Conversations of students on social networking sites like twitter, facebook throw light on education experiences like emotions, concerns. Twitter is a micro-blog with each tweet within 100-150 words so we can understand emotions of candidates. Most tweets are related to emotions, which tweets fall under which emotion. In this paper we are focusing to develop a model which predicts student’s emotion and understand their feelings, opinions related to their educational experiences. Few labels which we have used for fetching the tweets related to students are exams, results, engineering. Main phases in this application are text cleaning, processing, validation and prediction. In pre-processing/cleaning phase stop-words removal, stripping white-space, removing punctuation. In processing phase, document term matrix, creating corpus and applying supervised learning paradigms on training data. We validated the accuracy of model using 5-fold cross validation in validation phase. On the basis of training data, predicted the label of tweets in test data.