• 

  Interface

Smart Advice Box — Hackathon Project

Inspiration

Most virtual assistants can understand what you say, but very few truly understand how you feel — especially in non-English languages like Arabic.
I wanted to build something that can listen to your emotions and give meaningful responses — not just information.

As an Arabic speaker, I also wanted to build something inclusive, that supports both Arabic and English equally well.

What it does

Smart Advice Box is a bilingual emotional assistant that listens to users in Arabic and English, detects their emotional state, and responds with empathetic, intelligent advice rather than mere factual information.
It supports both voice and text inputs, delivering multi-language voice responses that feel natural and human.

How we built it

• Used Hugging Face sentiment analysis models:
  
  • MARBERT for Arabic emotion detection
    
  • DistilBERT for English sentiment detection
    
• Built a Flask API to route requests to the appropriate model based on the detected language
  
• Integrated the backend with Bolt.new, which powers the voice-enabled user interface
  
• Added multi-language voice responses using ElevenLabs
  
• Created a simple frontend supporting both voice and text input methods

Challenges we ran into

• Ensuring accurate emotion detection on short user inputs (e.g., "I'm sad" / "أنا متضايق")
  
• Fixing repeated voice playback issues in the Bolt interface
  
• Supporting reliable, seamless handling of both Arabic and English languages within the same system
  
• Building, testing, and deploying the entire solution in under 24 hours

Accomplishments that we're proud of

• Delivering a fully functional bilingual emotional assistant in less than 24 hours
  
• Seamlessly combining multiple NLP models and voice technologies for an inclusive user experience
  
• Creating a responsive voice and text UI powered by AI models
  
• Successfully handling the nuances of Arabic emotion detection alongside English sentiment analysis

What we learned

• How to integrate language detection with multilingual NLP models in one system
  
• How to implement multi-language emotional logic with fallback mechanisms
  
• How to rapidly build, test, and deploy AI-powered interfaces using Bolt.new
  
• How to fine-tune emotion-based user experience to feel more human and empathetic

What's next for Smart Advice Box

• Improve emotion detection accuracy, especially for subtle or ambiguous user inputs
  
• Expand support to additional languages and dialects beyond Arabic and English
  
• Enhance the user interface with richer conversational capabilities
  
• Integrate more advanced AI models for deeper emotional understanding and advice personalization
  
• Explore deployment on multiple platforms to increase accessibility

Built With

• ai
• bolt.new
• javascript
• lucide
• react
• tailwind
• typescript
• vite