Abstract:

Globalization and the proliferation of Massive Open Online Courses (MOOC) has radically altered the model of education. New technology in this field offers the opportunity to increase the availability of courses to a far greater audience than that provided in the traditional setting. However, the implementation has significant challenges that must be overcome to allow students to take full advantage of them. The flexibility of the platforms in which MOOC operate, and the large amount of learning resources they provide, allows for the inclusion of large numbers of students across a greater geographical base. This interaction between students and systems produces massive learning behaviour data and leaves traces of the educational process on systems that are useful for analysis. This amount of information presents a series of challenges that are difficult to manage through the use of traditional methods. Educational data mining has become an important discipline for discovering new and potentially useful information on large amounts of data. The implementation of these methods and tools allows the availability of statistical data that can contribute to the identification of the level of commitment and motivation that students have in this type of courses.

In this article, we study the analysis of sentiment of an MOOC, through the application of text mining techniques on messages received in the social network Twitter. Given the large volume of tweets that are generated around a MOOC, it is convenient to develop methods that are oriented to the processing of texts automatically with an acceptable accuracy. The purpose of this study is to analyze students' opinions about their courses, their instructors, and the main tools used on the course. Interactions between students and MOOC can be explored using text mining techniques, with the aim of improving learning and personalizing the students' experience.

This can be met by:
(1) predicting the level of popularity of the courses;
(2) obtaining feedback on the content of the courses so that the tutors can analyse and improve their teaching strategies;
(3) obtaining feedback on the platform support so that administrators can improve user experiences.

The analysis focus on the calculation and analysis of the frequency of terms, the analysis of concordances, groupings and n-grams. In addition a model was made for classification of tweets according to their polarity of sentiment, based on the algorithm used in machine learning known as Support Vector Machine (SVM). This work uses as an case study a MOOC hosted in Google in the “Actívate” platform in collaboration with the University of Alicante Spain.

Keywords:

Sentiment analysis, Opinion mining, Massive Open Online Course, MOOC, Supervised learning, Text mining.