LSTM built using PyTorch to predict whether a certain review is positive or negative (sentiment).

🍟 Results

A noticeable plateau occurs at around Step 5000 at the first epoch. A decrease in alpha (learning rate) is recommended and an increase in the number of embedding dimensions and hidden dimensions.

Test Loss: 0.211
Test Accuracy: 91.692%

🍰 Training the Model

• Download the dataset
• Extract the dataset
• Run the jupyter notebook

⚙️ Requirements

• Python >= 3.7.0
• Jupyter Notebook
• NumPy
• PyTorch
• CUDA 11.0
• Optional: tqdm

📜 Dataset

Amazon Reviews for Sentiment Analysis (Kaggle)

🧑 Author

• Kristian Espina