Multilingual Classroom Doubt Normalizer

nspiration

In classrooms, many students hesitate to ask doubts or express them poorly, especially when using regional languages like Tamil or mixed English. These vague or emotional questions make it difficult for teachers to understand the actual academic intent. This gap in communication inspired us to build a system that helps students express their doubts clearly without fear or confusion.

What it does

Multilingual Classroom Doubt Normalizer takes a student’s poorly framed doubt in Tamil, Tanglish, or English and converts it into a clear, concise, teacher-ready academic question. The system identifies the subject, extracts the student’s intent, and restructures the doubt into a formal academic format. If essential information is missing, it asks exactly one clarification question.

How we built it

We built the project using Google AI Studio with Gemini 3. Gemini 3 is used for multilingual understanding, intent extraction, subject detection, and controlled question normalization. A strict prompt enforces academic tone, prevents answering or teaching, and limits clarification to a single question. The application is implemented as a simple web-style interface inside AI Studio for fast testing and demonstration.

Challenges we ran into

The main challenge was preventing the model from answering or explaining the doubt instead of restructuring it. We also had to carefully control verbosity and ensure that clarification questions were asked only when absolutely necessary. Handling code-mixed Tamil-English input consistently was another challenge that required prompt refinement.

Accomplishments that we're proud of

Successfully normalized vague, multilingual student doubts into teacher-ready questions

Achieved reliable subject detection and intent extraction

Implemented strict single-clarification logic without spam

Built a clean, usable demo entirely with Gemini 3 in a short time

What we learned

We learned how powerful prompt design and constraint-based generation can be when using large language models like Gemini 3. We also learned the importance of focusing on a specific real-world problem and solving it cleanly instead of overloading features.

What's next for Multilingual Classroom Doubt Normalizer

In the future, this system can be integrated into classroom portals, learning management systems, or voice-based interfaces. It can also be extended to support more regional languages and provide analytics on common student confusion areas without replacing the teacher’s role.

Built With

• 3
• ai
• css3
• gemini
• google
• html
• javascript
• prompt-based