README.md

metadata

title: How Am I Feeling? - Emotion Classifier
emoji: 🎭
colorFrom: purple
colorTo: blue
sdk: docker
pinned: false
license: mit

🎭 How Am I Feeling? - AI Emotion Classifier

An AI-powered emotion detection system that analyzes text and identifies 10 different emotions with 88.6% accuracy.

🌟 Features

• 10 Emotion Classes: happiness, sadness, fear, embarrassment, disgust, drive, surprise, loneliness, love, excitement
• Beautiful Web UI: Modern, responsive interface with real-time analysis
• High Accuracy: 88.6% validation accuracy
• Fast Inference: <10ms per sentence
• Word2Vec + Neural Network: 300-dim embeddings → 128→64→10 network

🚀 Try It Out

Simply type or paste any text to see what emotions it conveys!

Example sentences:

• "I'm so grateful for this beautiful day!" → happiness
• "I miss the way things used to be" → sadness
• "I can't wait for the concert tomorrow!" → excitement
• "Every moment with you makes my heart complete" → love
• "I'm terrified of what might happen next" → fear

🧠 Technical Details

Architecture

Input Text → Preprocessing → Word2Vec (300-dim) → Neural Network (128→64→10) → Top-5 Predictions

Dataset

• Size: 100,000 sentences (10,000 per emotion)
• Source: Generated using LLaMA 3.1 70B via Deepinfra
• Quality: Diverse, natural language examples

Model

• Embeddings: Word2Vec (Skip-gram, 300 dimensions)
• Classifier: Fully-connected neural network
• Parameters: 34,634 trainable parameters
• Training: 50 epochs with early stopping
• Validation Accuracy: 88.6%

📊 Performance

Per-emotion accuracy:

• Best: happiness, love, excitement (~92%)
• Good: sadness, fear, surprise (~88%)
• Moderate: embarrassment, drive, disgust (~84%)

💻 API Usage

curl -X POST http://your-space-url/analyze \
  -H "Content-Type: application/json" \
  -d '{"text": "I am so excited about this!"}'

Response:

{
  "success": true,
  "predictions": [
    {"emotion": "excitement", "confidence": 0.92, "percentage": 92.0},
    {"emotion": "happiness", "confidence": 0.85, "percentage": 85.0},
    ...
  ]
}

🛠️ Built With

• TensorFlow/Keras - Deep learning
• Gensim - Word2Vec embeddings
• Flask - Web framework
• NLTK - Text processing

📝 License

MIT License - Free to use for personal or commercial projects!

🔗 Links

• GitHub: emotion-classifier
• Dataset: Coming soon to Hugging Face Datasets

---

Built with ❤️ and Python