Work from one of our previous CSER student projects, led by Katrina Le and Hamid Tarmazdi, has been accepted for publication at Latice 2017. This work explores how text processing techniques can be used to assist educators in identifying students who need assistance, by identifying unanswered questions, or questions that are expressing strong emotion, and forms part of a larger series of work in learning analytics for forum analysis.
Katrina Le, Hamid Tarmazdi, Rebecca Vivian, Katrina Falkner, Claudia Szabo and Nickolas Falkner, Directing Teacher Focus in Computer Science Online Learning Environments. Accepted for Latice 2017.
Discussion forums play a key role in most university courses today as a way to provide support for students outside classroom hours. However, with large class sizes and growing workloads for academics, monitoring often-large discussion fo- rums is not an easy task. As a result, situations where students are distressed, questions are unanswered, or students require urgent support, may go unnoticed. Text classification and sentiment analysis techniques have become a popular approach to determine user attitudes, emotions and experiences within business and social science domains. Initial research has begun to explore the application of text classification to students written text to investigate how students experience learning processes. In this paper, we build on this emerging field, and apply text classification to forum text to determine if we can correctly notify lecturers when a student is experiencing difficulties with their Computer Science studies. We implement a Nave Bayes Classifier and apply it to a Moodle forum data. Our results show the potential benefits of this approach and also highlight key avenues for future work.