Cookies

We use cookies to ensure that we give you the best experience on our website. You can change your cookie settings at any time. Otherwise, we'll assume you're OK to continue.

Durham University

Computer Science

Profile

Publication details for Mrs Laila Alrajhi

Alrajhi, L.M. (2019), Classification of Instructor Intervention in MOOC Environment, Early Career Researcher Conference 2019. Liverpool.

Author(s) from Durham

Abstract

Massive Open Online Courses (MOOCs) are one of the latest initiatives in open education. Their platforms contain many courses on different subject domain. For each such course, there are thousands of students and their comments to each part of the course, and it is difficult for the instructor to answer to all of them. However, the retention on such MOOC courses is low (in average, around 10%). Thus, it would be extremely useful for instructors to know which comments to answer, to encourage retention.
The main objective of this research is thus to build a Classification Machine Learning Model to predict the instructor intervention need with more accuracy. The comment, which is written by a student, is classified into three categories (Urgent intervention, Non-Urgent intervention, No intervention). The model will be trained based on linguistic features and Natural Language Processing (NLP) techniques. Different algorithms will be applied from Machine Learning (ML) for example (Logistic Regression, Naive Bayes, SVM, Random Forest and Gradient Boosting), Neural Networks (NN) and Deep Learning for example (Convolutional Neural Network, Recurrent Neural Network – LSTM, Recurrent Neural Network – GRU, Bidirectional RNN and Recurrent Convolutional Neural Network). In ML, different NLP features will be extracted, then they will constitute the input for training the model with different ML and NN algorithms. Then, the next step is to compare all these models and select the best one to predict the type of instructor interventions. The final stage is to make the model more intelligent, by examining more features (prior knowledge), such as, the number of likes and sub-comments, in addition to NLP features, to increase the prediction accuracy in identifying when a comment needs instructor intervention.