KAR³L: A Fresh Approach to Flashcards

Flashcards are effective tools to help students learn a variety of subjects, such as vocabulary, trivia, and law. Flashcard learning software like Anki and Quizlet are highly popular among users, but their underlying algorithms overlook a crucial part of flashcards: the textual content on the cards!

To effectively utilize the rich textual information present in flashcards, we've designed KAR³L, the first NLP-powered flashcard scheduler. The system behind KAR³L leverages pretrained language models and neural networks to more accurately predict which flashcards a student can answer correctly, informing the order in which flashcards are shown to the student. Recently, we've also been exploring how large language models can help generate study materials like mnemonic devices to help students learn more effectively. Come study with KAR³L and help us build the next generation of NLP-powered adaptive learning tools!

• Matthew Shu (Yale University)
• Nishant Balepur (Univeristy of Maryland)
• Shi Feng (George Washington Univeristy)
• Jordan Boyd-Graber (University of Maryland)

• Matthew Shu, Nishant Balepur, Shi Feng, Jordan Boyd-Graber. KARL: Knowledge-Aware Retrieval and Representations aid Retention and Learning in Students. arXiv preprint arXiv:2402.12291 2024

Description

Exams like the Graduate Record Examination (GRE) often encourage and require the rote memorization of hundreds or even thousands of challenging vocabulary words. While flashcard schedulers can aid this process, research has also shown that mnemonic devices can help make vocabulary learning more effective and engaging. For example, to learn the definition of "benevolent", you might remember a short mnemonic device like "Benevolent sounds like benefits, and someone who gives benefits is kind." Our goal is to study whether we can train a model optimized to generate mnemonic devices that help students, specifically GRE test-takers, learn vocabulary.

If you study our GRE Vocab deck in KAR³L, you will be participating in our research study, which spans from 02/26/2024 to 05/26/2024. In this research study, we will be collecting your feedback to improve our mnemonic device generation. None of your personal information will be collected or released. Users participating in our study are eligible to receive monetary rewards, summarized below. Full instructions for our user study can be found here.

Monetary Reward Structure

As a token of appreciation for helping us with our study, we will have two ways for users to win rewards (both rewards can be earned):

1. Base Reward: Users who study 20+ GRE flashcards a day over 5 unique days will be entered into a raffle to win one of 50 $25 gift cards
2. Power User Reward: Users who give feedback on 75 or more mnemonic devices will be entered into a raffle to win one of 50 $25 gift cards

You can check out the "statistics" page to see your progress on these rewards! (see "Vocab Facts Studied" and "Mnemonics Rated")

Contact

If you have any questions, please contact me (Nishant Balepur), a co-investigator of the project (University of Maryland, nbalepur@umd.edu). You can also join our Discord. Thanks again for your help, and happy studying!

Acknowledgments

Any opinions, findings, and conclusions or recommendations expressed in this material are those of the researchers and do not necessarily reflect the views of the sponsor.

<< back to top