🧠 AI-Powered Multi-Class Text Classifier
Production-Ready NLP Pipeline & Deployment
A sophisticated, end-to-end Natural Language Processing (NLP) system designed to categorize unstructured text into meaningful classes. This project bridges the gap between raw data and actionable intelligence using a combination of classical Machine Learning and state-of-the-art Deep Learning.
🚀 Key Features
- Multi-Model Support: Compare results between Logistic Regression (TF-IDF) and Deep Learning (LSTM) architectures.
- Advanced Preprocessing: Implements POS-aware lemmatization, tokenization, and stopword removal via a modular vectorized pipeline.
- Smart Embeddings: Supports Word2Vec Skip-gram and GloVe representations for deep contextual understanding.
- Premium UI: A modern, responsive web interface built with Glassmorphism and high-end typography for instant narrative analysis.
- Scalable Architecture: Designed to handle datasets with millions of rows while maintaining high performance.
🛠️ Tech Stack
- Core: Python 3.8+
- NLP: NLTK, Gensim
- Machine Learning: Scikit-Learn (TF-IDF, Logistic Regression, Random Forest)
- Deep Learning: TensorFlow/Keras (LSTM, GRU, Word2Vec)
- Deployment: Flask (Backend), Modern Vanilla CSS (Premium Frontend)
📂 Project Structure
├── Datasets/ # Training & Test CSV files (Standard: 'QA Text' & 'Class')
├── models/ # Exported model "brains" (pkl, h5, tokenizer)
├── templates/ # HTML templates for the Flask web application
├── text_pipeline.py # The core NLP engine (preprocessing & utilities)
├── train_script.py # High-speed training script with Stratified Sampling
├── train.ipynb # Interactive research & experimentation notebook
├── app.py # Flask deployment server
└── requirements.txt # Project dependencies
🏁 Quick Start
1. Installation
Clone the repository and install the necessary libraries:
pip install -r requirements.txt
2. Training the "Brain"
You can train the models either interactively via the notebook or quickly via the terminal:
# High-speed stratified training
python train_script.py
3. Launching the Dashboard
Once the training is complete and artifacts are in the models/ folder:
python app.py
Visit http://127.0.0.1:5000 to start analyzing text!
🧪 The "Chaotic Text" Test
The system is built to handle noise. Try pasting a "chaotic" sample like this to see how the model ignores noise and finds the core intent:
"MARKET ALERT! 📉 Stocks are tumbling as investors react... Should I diversify into gold?? #WallStreet #Investing"
📄 License
This project is licensed under the MIT License - see the LICENSE file for details.
Built with ❤️ for High-Performance NLP.