The inspiration for TaleTutor came from the desire to make learning more engaging and effective for students. We recognized that traditional educational methods often fail to connect topics holistically and relate concepts to real-world scenarios, leading to disinterest and fragmented understanding. Drawing from research on narrative learning and the transformative power of storytelling, we envisioned an AI-driven platform that uses immersive narratives to captivate students' imaginations. TaleTutor strives to bridge the gap between abstract concepts and practical applications, making education both fun and meaningful.

TaleTutor is a narrative-based learning system that serves three primary purposes:

1. Facilitating Key Connections: Enabling students to make essential connections between concepts by utilizing the power of storytelling.

2. Relating Material to Real-World Scenarios: Connecting academic content with practical applications fosters context-based learning.

3. Making Learning Enjoyable: Using immersive narratives to make education fun.

1. Dual Interfaces: Separate interfaces for teachers and students.

2. Themed Learning: Students can choose their learning theme, such as Harry Potter, with subjects like Physics and Civics.

3. Interactive Chat System: A character-led interactive chat system that dynamically explains concepts based on student interaction.

4. Real-World Scenario Teaching: Using fictional characters to teach real-world scenarios through engaging narratives.

TaleTutor utilizes a LangChain-supported GPT-3.5 LLM to retrieve PDF content uploaded by teachers, convert it into a vectorstore, and perform RAG operations based on student queries. The system processes RAG output with another GPT-3.5 LLM that acts as a router, deciding whether to proceed with narrative building or knowledge retrieval. The GPT-3.5 LLM then generates narratives using zero-shot prompting. The system also manages off-topic queries by guiding students back on track or notifying teachers for relevant follow-up.

Tech Stack:

• GPT-3.5 LLM: For generating and routing narratives.

• LangChain: For managing RAG operations.

• Flask: Python server to interface with LLMs.

• Next.js: Frontend framework for the user interface.

• Axios: Library for making HTTP requests to the server.

• TaleTutor was selected as one of the top 10 projects in UC Berkeley's AI Hackathon 2024 out of 290 projects.

• TaleTutor prevents students from going off-topic, avoids hallucinations, and creates tickets for teachers to address out-of-syllabus questions.

• TaleTutor accurately integrates lessons with movie stories, ensuring distinguishable custom voices for movie characters.