Predicting Responses to Spaced Repetition Flash Cards with Machine Learning Techniques

Thesis Report Download (PDF) Online Learning Software

Abstract:

With the development of personal computers and smart phones, many new learning environments and learning methods are emerging. With this, educational data collection has become trivial and is becoming a powerful tool for improving education. This project created an online flash-card environment for memorising foreign vocabulary, and attempted to predict whether a student will be able to recall a particular word in a target language by tracking, recording and analysing reviews with machine learning techniques. This has many implications in education, including identifying struggling students, identifying difficult vocabulary to focus on, as well as improving methods of memorisation of foreign vocabulary. The web based software was available from desktop and mobile devices and allowed students to study anywhere. Over 13 weeks, 28 students completed over 7,500 reviews. The results were promising, with the system correctly predicting student recall of approximately 70% of flash-card reviews. The rate of forgetting was also plotted based on the review data, however the amount of data and number of reviews was too small for any significant findings. Future work might involve a larger number of more motivated students to gather a larger dataset.