WikiFlash: Generating Flashcards from Wikipedia Articles

Abstract

Flashcards, or any sort of question-answer pairs, are a fundamental tool in education. However, the creation of question-answer pairs is a tedious job which often defers independent learners from properly studying a topic. We seek to provide a tool to automatically generate flashcards from Wikipedia articles to make independent education more attractive to a broader audience. We investigate different state-of-the-art natural language processing models and propose a pipeline to generate flashcards with different levels of detail from any given article. We evaluate the proposed pipeline based on its computing time and the number of generated and filtered questions, given the proposed filtering method. In a user study, we find that the generated flashcards are evaluated as “helpful” on average. Further, users evaluated the quality of human created flashcards that are available open source as comparable to or only slightly better than the automatically generated cards.