AI Study Buddy: An Automated NLP Pipeline for Generating Active-Recall Flashcards to Reduce Cognitive Load

The difficulty of managing an overwhelming amount of academic material within limited time constraints is familiar to most students. Although extensive cognitive science research demonstrates that active recall and spaced repetition are highly effective for information retention, the majority of students continue to use passive study methods such as rereading, highlighting and note copying. This habit persists primarily because building a comprehensive flashcard deck from scratch is nearly as time-consuming as studying the material itself. This paper presents AI Study Buddy, a web application which automates the creation of study flashcards. Users can submit their academic notes, which the system processes to identify key concepts and generate question and answer flashcards for study. The application uses a five-stage natural language processing (NLP) pipeline: it segments documents into intersecting chunks, uses spaCy for concept extraction, leverages GPT-3.5-Turbo to generate questions at varying cognitive levels, produces answers strictly based on the reference text and applies cosine similarity scoring to detect and filter out hallucinated content. Flashcards with a similarity score under 0.80 are excluded from the final set. The front end of the application is developed using Next.js and React, while the backend operates on Python Flask. Data storage is managed by MongoDB Atlas. The system is deployed on Vercel and is freely accessible on https://the-flashcard.vercel.app. Testing has verified that all modules work as intended and that flashcards are generated within reasonable timeframes. No subscription is needed for access.