Squizzfoot

Inspiration

The inspiration behind this project stemmed from the growing demand for interactive and automated quiz-based learning platforms and to gather people passionnate about soccer games . Leveraging Reddits platform and AI technologies we wanted to offers soccer fans great quizzes about football facts generated by AI with both general knowledge and custom knowledge. This not only enhances learning experiences but also makes quizzes more dynamic and adaptable to various types of content. We chose to focus on football news articles to create an engaging and educational experience for football fans and to challenge their memories on bigger events.

What it Does

Our project is an quizz of multiple-choice questions based on Football our AI-powered quiz generation system automatically creates, such as football news articles. The AI reads the article, extracts relevant information, and then generates structured quiz questions along with possible answers. These questions are then stored in a sql database (postgres) for later use, creating a dynamic and automated quiz system that can be used to test knowledge or build educational content.

Key features:

• Automatically generates multiple-choice with AI.
• Organizes questions and answers in a structured format.
• Saves the questions and answers in a SQL database for future use or integration into learning systems.

How We Built It

We used several modern technologies and AI tools to bring this project to life:

• Langchain: We used Langchain to integrate language models like OpenAI's GPT (or Google Gemini) to generate questions based on the content of football news articles.
• PostgresSQL: For storing and managing the generated questions and answers, we interracted with the database through psycopg2.
• Python: We wrote Python scripts to interact with the language model, parse responses, and save the questions in the database.
• Regex: We used regular expressions to parse AI-generated responses and extract the questions, options, and correct answers in a usable format.

The entire process involves feeding a text article into the system, which then outputs a set of questions based on the content, parses them, and stores them for future use in quizzes.

Challenges We Ran Into

• Integrating Langchain with the AI Model: The integration of Langchain with a specific AI model was a bit challenging at first, particularly fine-tuning the prompt to generate specific types of questions like multiple-choice, true/false, and fill-in-the-blank questions. We had to experiment with different prompts and models to get accurate and relevant results.
• Parsing the AI Responses: Another challenge was parsing the free-form responses from the AI into structured formats that we could use in our SQL database. We needed to handle inconsistencies in the formatting of the generated questions and options.
• Ensuring Content Relevance: Since the system generates questions based on the provided content, we had to fine-tune how the AI models prioritize which parts of the article are most relevant for creating insightful questions. Getting high-quality questions with minimal noise took time and experimentation.

Accomplishments That We’re Proud Of

• Automated Question Generation: We successfully built a system that can automatically generate multiple types of quiz questions from arbitrary content, which is useful for educational platforms, quizzes, or games.
• Structured Database Integration: We built a robust backend to store and organize the generated questions in a SQL database, making it easy to retrieve and use them later.
• Customizable Question Formats: We developed a system that can generate multiple-choice, true/false, and fill-in-the-blank questions based on the content of the article, providing flexible quiz options.
• AI-Driven Content Interaction: By leveraging Langchain and powerful language models, we created an intelligent system that can adapt to a variety of topics and generate relevant questions on the fly.

What We Learned

• Language Model Tuning: We learned a lot about fine-tuning prompts for language models to generate useful, accurate, and contextually relevant questions. It's a delicate balance to ensure the model understands the type of question to generate.
• Data Parsing and Regex: We learned how to use regular expressions effectively to parse unstructured AI responses into structured, usable data. This was a key challenge and a major learning curve.
• Database Design: We gained valuable experience in designing and working with SQL databases to manage and store dynamic data like quiz questions and answers.

What’s Next for Squizzfoot

• Enhancing Question Types: Moving forward, we plan to add support for additional question formats, such as open-ended questions or more advanced question types based on the content.
• Broader Content Sources: We will expand the system to generate quiz questions from a wider variety of content types, including sports news, educational articles, and even books.
• Improved AI Integration: We plan to integrate more advanced AI models and fine-tune them further for even better question generation. We are also exploring the integration of more specific tools like Google Gemini for a more accurate and responsive AI system.
• User Interface: We want to build a web interface or an app where users can input articles, view generated quizzes, and track their progress in answering the questions.
• Multi-Language Support: We aim to add support for generating questions in multiple languages to make the platform accessible to a wider audience.

Built With

• fastapi
• html5
• postgresql
• python