A multiplayer quiz service where you can solve AI-generated questions with friends.
1. Convert PDF content into quizzes
Users upload a PDF file, which is parsed and vectorized. Based on the extracted content, an LLM generates customized quiz questions.
2. Create a quiz room
A room is created where users can join to solve the quiz together. Users can communicate using simple emojis and chat messages.
3. Real-time quiz solving
Once the game starts, users solve questions in real-time. Questions are presented sequentially, and when a majority of players have answered, the game moves to the next question after a short countdown. Scores are updated live based on correct answers. Light interactions through emojis are supported during the game.
4. Share and discuss answers
After the quiz ends, a final scoreboard is shown. Users can view and share each other’s answers and engage in discussion about them.
| Start Page | Main Lobby | Create Room |
| Join Room | Solve Quiz | View Answers |
| Overall Architecture |
| CI / CD |
| Team WeQuiz | ||||
|---|---|---|---|---|
| Kyuyeon Park(Team Lead) | Jaemin Shim | Keumjang Ahn | Woorim Kim | Junhyeong Bae |
| 🤖 | ⚙️ | ⚙️ | 🎨 | 🎨 |
| ML Engineer | Back-end Developer | Back-end Developer | Front-end Developer | Front-end Developer |
I implemented the end-to-end quiz auto-generation pipeline based on RAG (Retrieval-Augmented Generation), enabling question creation from document content. [ 👉 directory ]
This pipeline includes:
- Document Summarization: Parsing and chunking uploaded PDFs using LangChain
- Question Generation: Extracts key concepts from the document and uses retrieved chunks to generate contextually relevant quiz questions via LLM prompts
- Answer Evaluation: Assessing user answers against ground-truth answers using similarity scoring
| Quiz and Document Summarization Workflow |
2024-1 KMUCS Capstone Design | Copyright 2024. WeQuiz All rights reserved.