Joint Exploration of Negative Academic Emotion and Topics in Student-Generated Online Course Comments

Abstract

Currently, with the increasing advancement of interactive learning technologies in MOOCs, a large number of student-generated comments (SGCs) have been substantially produced with two primary emotions (positive and negative). The emotional orientations are typically related with specific learning topics or aspects discussed, which is of value to offer abundant academic feedbacks for teachers and developers. Especially, the negative emotion and topics can be exploited to get an in-depth insight of the problems and barriers encountered by learners in online learning. However, it is challenging to capture relevant details from unstructured SGCs. In this paper, we propose a generative probabilistic model that extends Sentence-LDA (SLDA), namely Emotion Topic Joint Probabilistic Model (ETJM), to explore negative opinions in terms of pairs of which we call emo-topic. The model first automatically extracts the sentences with the high negative emotion density (NED), and then incorporates emotion and topic together to explore negative emotional feedbacks towards topics. The experimental results show that learners extended some negative comments towards the issues about learning content, online assignments and certificates of courses. The summarization of these issues can be given back to teachers to regulate and improve the teaching methods, strategies and design of learning contents.