← All work
Case study · 07/08
GenQuest
University project generating quiz questions from book PDFs using T5, spaCy and BERT.
(01) Problem
Users uploaded book PDFs and selected chapters. The system needed to generate relevant quiz questions with answers and multiple-choice options from the text.
(02) Approach
- 01T5 question generator model drafts initial questions based on the selected chapter text.
- 02spaCy extracts entities to supply answers and construct multiple-choice options.
- 03BERT QA evaluator ranks the question-answer pairs to filter results.
- 04React front end displays the generated questions alongside an Express API for account management.
(03) Architecture
PDF UploadT5 GeneratorspaCy EntitiesBERT EvaluatorReact Frontend
(04) The hard part
The hardest part was ensuring the generated questions were accurate and relevant to the text. This was handled by using a BERT QA evaluator to rank pairs and retain only the best ten results.
(05) Result
The system successfully generated quiz questions with answers, such as identifying lunar sample weights.