Sentiment Classification on IMDb Dataset

This project involves sentiment classification on the IMDb dataset using two different models: an LSTM and a simple Transformer. The goal is to classify movie reviews as positive or negative.

Models

LSTM

• Accuracy: 90% after 10 epochs

Transformer

• Accuracy: 80% after 20 epochs

Results

LSTM Confusion Matrix	Transformer Confusion Matrix

LSTM ROC Curve	Transformer ROC Curve

Dependencies

The project uses the following libraries:

• torch
• torchvision
• torchtext
• numpy
• matplotlib
• sklearn

Installation

To install the required dependencies, run:

pip install -f requirements.txt

Usage

1. To choose model, see the config
2. Train

python train.py

3. Evaluate/visualize

python evaluation.py