🎭 Emotion Classification System

A comprehensive machine learning-powered emotion classification system that analyzes text and predicts emotions with confidence scores. Built with Python, scikit-learn, and Streamlit for an interactive web interface.

🌟 Features

• Multi-emotion Classification: Detects various emotions including joy, sadness, anger, fear, disgust, surprise, and more
• Interactive Web Interface: Beautiful Streamlit-powered web app with real-time analysis
• High Accuracy: Achieved best performance with Logistic Regression + Count Vectorizer (LogReg_Count)
• Comprehensive Analysis: Provides confidence scores for all emotion categories
• Text Processing: Advanced NLP preprocessing with stemming, lemmatization, and stop word removal
• Visualization: Dynamic charts and word clouds for emotion distribution
• File Upload Support: Analyze text from uploaded files
• Text Statistics: Detailed analysis of input text characteristics

🚀 Demo

Web Interface

The Streamlit app provides an intuitive interface for emotion analysis:

• Text Input: Type or upload text files for analysis
• Real-time Results: Instant emotion prediction with confidence scores
• Visual Analytics: Interactive charts showing confidence distributions
• Text Statistics: Comprehensive text analysis metrics

Sample Predictions

Input: "I'm so excited about this new opportunity!"
Output: JOY (87.3% confidence)

Input: "This situation makes me really angry and frustrated."
Output: ANGER (92.1% confidence)

📊 Model Performance

Our emotion classification system was trained and evaluated on multiple algorithms:

Model	Vectorizer	Accuracy
Logistic Regression	Count Vectorizer	Best Performance ⭐
Logistic Regression	TF-IDF	High Performance
Naive Bayes	Count Vectorizer	Good Performance
Naive Bayes	TF-IDF	Good Performance

Key Metrics

• Best Model: Logistic Regression with Count Vectorizer
• Text Preprocessing: Stemming, Lemmatization, Stop word removal
• Feature Engineering: Count-based vectorization
• Evaluation: Comprehensive metrics including accuracy, precision, recall, and F1-score

🎯 Usage

Running the Web App

1. Start the Streamlit server

   streamlit run app.py

2. Access the application

   • Open your browser and go to http://localhost:8501
   • The app will automatically load the trained models

🔬 Technical Details

Data Preprocessing Pipeline

1. Text Cleaning

   • Convert to lowercase
   • Remove punctuation and special characters
   • Filter out non-alphabetic tokens

2. Tokenization

   • Split text into individual words
   • Remove stop words (common words like 'the', 'and', etc.)

3. Normalization

   • Lemmatization: Reduce words to their root form
   • Stemming: Remove suffixes to get word stems

4. Vectorization

   • Convert text to numerical features
   • Count Vectorizer for best performance

Model Architecture

# Best performing pipeline
Pipeline([
    ('vectorizer', CountVectorizer()),
    ('classifier', LogisticRegression(random_state=0))
])

Supported Emotions

The system can classify the following emotions:

• 😊 Joy
• 😠 Anger
• 😨 Fear

📈 Visualization Features

Training Visualizations

• Emotion distribution charts
• Word clouds for each emotion category
• Model performance comparison
• Confusion matrices for each model

Web App Visualizations

• Real-time confidence score charts
• Interactive emotion distribution
• Text statistics dashboard
• Dynamic progress indicators

🚀 Deployment

Local Deployment

streamlit run app.py --server.port 8501

📖 API Documentation

Core Functions

preprocess_text(text)

Preprocesses input text for emotion classification.

Parameters:

• text (str): Raw input text

Returns:

• str: Cleaned and processed text

predict_emotion(text, model, vectorizer)

Predicts emotion for given text.

Parameters:

• text (str): Input text to classify
• model: Trained classifier model
• vectorizer: Fitted text vectorizer

Returns:

• tuple: (predicted_emotion, confidence_scores)