Inspiration
We wanted to personalize and humanizing online teaching.
What it does
This project is an AI-powered educational tool that creates personalized quizzes from any given text document. It uses facial sentiment analysis to gauge the user's mood, making the quiz generation process more empathetic and human-like. For example, if the user appears frustrated, the system intelligently adjusts the difficulty level, providing easier questions to enhance the learning experience and reduce stress.
How we built it
Our project, an innovative AI-powered educational tool, was meticulously built using Flask as the backbone for our backend infrastructure. The system is designed to generate personalized quizzes from any given text document, with a unique feature of adjusting the difficulty level based on the user's mood and performance.
The quiz questions are generated by querying the OpenAI API. To enhance the quality and relevance of these questions, we integrated Hume, a sentiment analysis tool. Hume's role is to gauge the user's emotional state, particularly feelings of sadness or frustration, and inform the system. Based on this input, we then craft a more empathetic and emotion-aware prompt to query the OpenAI API, ensuring the generated questions are sensitive to the user's current emotional state.
The personalization aspect of our system is further enhanced by the adaptive difficulty feature. If the user is consistently answering the questions correctly, the system gradually increases the complexity of the subsequent questions. Conversely, if the user is struggling with the questions, the system reduces their difficulty level. This dynamic adjustment ensures a tailored and effective learning experience for each user, promoting engagement and reducing frustration.
Challenges we ran into
We were originally using the llama_index API to do the quiz generation. However, we found it difficult to set the role of the query so we switched to the OpenAI API which made it easier to query as different user hierarchies. We also had some difficulties with inconsistencies in sending information to the Hume API. It would sometimes work and sometimes not work. We still don't know what the problem is. We had many challenges with contradictory python dependencies. We also had a lot of issues with setting our github up and having a different remote origin fetch than our origin main. So we we're always fetching from the wrong repository.
Accomplishments that we're proud of
- Generating quizzes and being able to display it.
- Using Hume to do live facial sentiment analysis ## What we learned We learned about virtual environments in Python We learned how to use APIs and the high-level model on how ChatGPT is implementing ## What's next for EmpathyQ
Log in or sign up for Devpost to join the conversation.