Sentiment Analysis on Massive Open online Courses (moocs)



Sentiment Analysis on Massive Open Online Courses (MOOCs)

Aswin S Prabhakar, Rajeev EM, Renny George Babu, Sudheesh P, Jasmine T Bhaskar

Dept. of Computer Science and Enginering Amrita School of Engineering, Kollam, India

{aswinsprabhakar, betakid007, rennygeorge07, sudheeshperumbil} @gmail.com, jasmine@am.amrita.edu

Abstract—This paper reports on the results of an analysis on user reviews of MOOCs. Using data collected from http://www.coursetalk.com, a leading source for student-powered reviews MOOCs, we determine whether a given review is positive or negative. By applying the Naïve Bayes and Support Vector Machine (SVM) methods, we find that standard machine learning techniques definitely outperform human-produced baselines. We conclude by examining factors that make the sentiment analysis problem more challenging.

Index Terms—Massive Open Online Courses, Naïve Bayes, Sentiment Analysis, SentiWordnet, Support Vector Machines, User Reviews

I. INTRODUCTION

M

ASSIVE open online courses are a recent addition to the range of online learning options. Many academics have taken interest in MOOCs recognizing the potential to deliver education around the globe on an unprecedented scale. MOOCs are widely discussed across a range of media, including blogs and the specialist and popular press. A MOOC brings together people interested in learning (or “students”) and an expert or experts who seek to facilitate the learning. Connectivity is usually provided through social networking, and a set of freely accessible online resources provides the content or the study material. Participation in a MOOC is completely voluntary and is dependent on the interested individual. More recently MOOCs have developed within international co-operative partnerships such as Coursera (www.coursera.org), a partnership of 62 world class universities (as of April 11, 2013) led by Stanford University, and edX (www.edx.org) which includes the Massachusetts Institute of Technology, École Polytechnique Fédérale de Lausanne, The Hong Kong University of Science and Technology. These two form the most comprehensive source

for the data that was extracted for further processing.

MOOCs raise a number of interesting challenges when doing research, such as measuring participation, and defining success. However, there has been a rapid growth in online discussion forums and review sites, where a crucial characteristic of the posted articles is their sentiment, or overall opinion towards the subject matter.