Improving Learn(ing): Quizlet’s New Memory Model

Quizlet’s data science team works behind the scenes to improve the Quizlet study experience by applying techniques from the fields of machine learning and cognitive science. You might have seen our last blog post about how we order terms in Learn mode using a Multiple Logistic Regression model based on previous answer history and time elapsed between answers. Our next step was to take the data from launching Learn to improve how we choose which terms you should study next.

Learn is now one of Quizlet’s most popular study modes, with almost a million users per day and over 250 million study sessions logged to date (only Flashcards are more popular!). All improvements we make to this study mode have an enormous impact on helping students efficiently practice and master what they’re learning.

This fall, we hosted a technical meetup at Quizlet HQ and presenters from all backgrounds and industries talked about how they build platforms for scientific research. One of our data scientists, Shane Mooney, presented a new and improved memory model called the Recurrent Power Law model and the practical considerations of implementing it while supporting offline capabilities. To learn more, catch the video of his talk here.

Interested in learning more about how we use data to power student learning? Catch us at the NIPS 2017 Teaching Machines, Robots, and Humans workshop where we’ll be presenting this work!