Regression-Chatbot

This project is a simple neural-network based chatbot built using PyTorch upon the machine learning concept of Linear Regression.
The chatbot learns to classify user input into predefined intents using a Bag-of-Words model and a small feed-forward neural network.

---

Features

• ✅ Written in Python with PyTorch
• ✅ Uses Bag-of-Words for text representation
• ✅ Neural network with 2 hidden layers (ReLU activation)
• ✅ Predicts user intent using softmax confidence score
• ✅ Simple training on hard-coded patterns
• ✅ Interactive chat mode in terminal

How It Works

1. The chatbot has predefined intents like greeting, goodbye, thanks, etc.
2. Each intent contains:
   • patterns (training sentences)
   • responses (what bot replies)
3. Text patterns are tokenized and converted to Bag-of-Words vectors.
4. These vectors are fed to a feed-forward neural network.
5. The model outputs probabilities → highest probability = predicted intent.

---

Technologies Used

Component	Library
Deep Learning Framework	PyTorch
Numerical Computation	NumPy
Model Training	Adam Optimizer + CrossEntropy Loss