Emotion Detection & Movie Recommendation Flask App

Project Description

This project is a Flask-based web application that detects the emotion from user input text and recommends movies based on the detected emotion. The app uses a pre-trained machine learning model to analyze the input and leverages movie data to provide personalized recommendations.

Features

• Emotion Detection: Enter any text, and the app predicts the underlying emotion.
• Movie Recommendations: Based on the detected emotion, the app suggests a curated list of movies.
• User-Friendly Interface: Simple web interface for input and displaying results.
• Dynamic Results: Movie details including title, release date, overview, and poster are shown.

How It Works

1. The user enters text on the homepage.
2. The backend predicts the emotion using a trained model.
3. The app fetches movies related to the emotion and displays them on a results page.

Technologies Used

• Python 3
• Flask
• joblib (for loading ML models)
• HTML/CSS (for frontend)
• dotenv (for environment variables)

Getting Started

1. Clone the repository
2. Create a virtual environment

   python -m venv venv
   

3. Activate the virtual environment
   • On Windows:

     venv\Scripts\activate
     

   • On macOS/Linux:

     source venv/bin/activate
     

4. Install dependencies

   pip install -r requirements.txt
   

5. Run the app

   python app.py
   

6. Open your browser and go to

   http://127.0.0.1:5000/
   

Folder Structure

• app.py - Main Flask application
• templates/ - HTML templates (home.html, result.html)
• model/ - Pre-trained emotion detection model
• API.py, test_modedel.py, etc. - Supporting modules

License

This project is for educational purposes.

ADDITIONAL INFORMATION

Technical Implementation:

• Designed and trained Multinomial Naive Bayes classifier from scratch achieving 97% accuracy on 1,000-sample emotional text corpus

• Engineered complete ML pipeline: text preprocessing → TF-IDF vectorization → model training → systematic evaluation → deployment

• Implemented comprehensive model evaluation: precision scores of 1.00 for 6/8 emotion classes and 0.88 for nuanced emotions (sad/lonely)

• Built Flask REST API with real-time emotion detection, external movie database integration, and dynamic content delivery

• Built functional web interface with form input validation and grid-based movie results display

Performance Metrics: Model Accuracy: 97% | Weighted F1-Score: 0.96 | Inference Time: <100ms Dataset: 1,000 balanced emotional text samples (125 per emotion category across 8 classes)

Technical Challenges & Solutions:

• Feature Engineering: Implemented TF-IDF vectorization with custom stopword removal and text normalization pipeline
• Model Selection Process: Benchmarked Multinomial NB against SVM and Logistic Regression; selected NB for optimal text classification performance
• API Integration: Designed RESTful movie database API integration with error handling, timeout management, and response caching Tech Stack: Python, scikit-learn, Flask, pandas, numpy, TF-IDF, RESTful APIs, HTML/CSS, joblib