Intent Classification with Feedforward Neural Network

Overview | Features | Requirements | Project Structure | Usage | Implementation Details | Dataset

Overview

Implementation of a Feedforward Neural Network for intent classification using only Python and NumPy. The model classifies user intents from text input using the Sonos NLU Benchmark dataset.

Features

• Bag-of-Words text representation
• Neural Network implementation with:
  • Input layer
  • Hidden layer with ReLU activation
  • Output layer with Softmax activation
• Training methods:
  • Batch gradient descent
  • Mini-batch gradient descent
  • Stochastic gradient descent

Requirements

• Python
• NumPy
• matplotlib (for plotting)

Install requirements:

pip install -r requirements.txt

Project Structure

assignment2/
├── data/
│   └── dataset.csv
├── model/
│   ├── __init__.py
│   ├── ffnn.py
│   └── model_utils.py
├── assignment2.py
├── utils.py
└── helper.py

Usage

1. Basic training with batch gradient descent:

python assignment2.py --train

2. Training with mini-batch/stochastic gradient descent:

python assignment2.py --minibatch --train

Implementation Details

• Learning rate: 0.005
• Number of epochs: 1000
• Mini-batch size: 64 (for mini-batch mode)
• Hidden layer size: 150 neurons
• Activation functions: ReLU (hidden layer), Softmax (output layer)
• Loss function: Cross-entropy

Dataset

Uses the Sonos NLU Benchmark dataset under Creative Commons Zero v1.0 Universal license.