exAImination

We always had a problem of trying to revising the chapters we have covered. So we came up with the idea of using AI to read the chapter/document and creating a set of questions to test yourself out.

We can upload a .txt file or add the text we read into the AI. This is reads the data and generates a set of questions according to the user's input. On answering these questions, the answers are then evaluated and then a score is generated.

We used OpenAI's APi and finetuned it to read the data and generate responses it needs. This generated series of questions and answers are sent to the Backend. The Backend then sends the questions as json object to the Frontend. The Frontend records the responses and sends it back to the Backend. The Backend evaluates it with the answer key it received. For the descriptive questions, the answer is evaluated upon the similarity score. This score is then shown to the user.

Challenges we ran into

1. Integration of different technologies: Integrating the AI model developed in Python with the Node.js backend and Next.js frontend seamlessly requires careful consideration of data formats, communication protocols, and API design.
2. Ensuring question quality and fairness: The generated questions should be clear, unambiguous, and grammatically correct. They should also be fair and unbiased, avoiding any cultural or demographic biases. This requires careful evaluation of the AI model's outputs and continuous improvement.

Accomplishments that we're proud of

1. Innovative Use of Technology: Combining Next.js for the front end, Node.js for the backend, and Python for model fine-tuning demonstrates a sophisticated understanding of different programming languages and frameworks. This combination allows for a seamless user experience, efficient data handling, and robust AI integration.
2. Automated Test Generation: The ability to automatically generate tests based on specific parameters and learning objectives saves educators and test creators significant time and effort. This automation also ensures consistency and quality in test creation.
3. Contribution to Educational Technology: Developing an AI-powered test maker contributes to the advancement of educational technology, providing educators with innovative tools to enhance student learning and assessment. This contribution has a positive impact on the overall education landscape.
4. Immediate Feedback and Score Reporting: Providing users with immediate feedback and score reporting after completing a test enhances their learning experience and allows them to gauge their understanding of the subject matter. This immediate feedback promotes self-assessment and motivates further learning.
5. Downloadable Questions and Answers: Offering downloadable questions and answers allows users to review their performance in detail and revisit specific concepts for further study. This feature provides a valuable resource for ongoing learning and test preparation.

What we learned

1. AI Integration and Model Fine-tuning: We learned how to integrate AI algorithms into a web application and fine-tune the model using Python. This experience provided insights into the practical application of AI in real-world scenarios.
2. Full-Stack Development Skills: We gained hands-on experience in full-stack development, combining Next.js for the front end, Node.js for the backend, and Python for model fine-tuning.
3. Collaboration and Teamwork: We developed collaboration and teamwork skills by working with team members with diverse expertise. This project allowed us to practice effective communication, problem-solving, and task delegation.
4. Continuous Learning and Improvement: We embraced the concept of continuous learning and improvement by refining our skills, seeking feedback, and incorporating new knowledge into the project. This mindset is crucial for success in the ever-evolving field of technology.

We can extend this to read pdfs by using "pdfjs-dist" module and this data can be read without needing to use text. This needs the Frontend to send the pdf to the Backend where the Backend processes it and sends the data to the finetuned model. We can also add the feature of downloading the questions with their answer key for teaching purposes

Built With

• css
• express.js
• fine-tuning
• javascript
• nextjs
• node.js
• openai
• python
• react
• tailman