👨💻 Author
Mr. Ritik Rana
Sentiment Analysis using IMDb Dataset
This project implements a Sentiment Analysis Model using the IMDb movie reviews dataset. The model is built using a custom pipeline in Keras Sequential API, with full control over tokenization and a deep learning architecture that includes Embedding and Bidirectional LSTM layers.
🧠 Model Architecture
The model is designed to classify movie reviews as either positive or negative. The architecture includes:
- Custom tokenization (manual control, not relying on Keras's built-in tokenizer).
- Embedding Layer: Converts integer-encoded words into dense vectors of fixed size.
- Bidirectional LSTM: Captures dependencies in both forward and backward directions.
- Dense Output Layer: With sigmoid activation for binary classification.
🗂 Dataset
- Source: IMDb Movie Reviews Dataset
- Classes: Binary (positive or negative sentiment)
- Preprocessing:
- Custom tokenization
- Padding of sequences to a fixed length
- Train-test split (commonly 80/20 or 70/30)
⸻
📊 Evaluation
• Accuracy: 96% (replace with your results)
• Loss: ~0.4 (replace with your results)
• Evaluation metrics: Accuracy
📌 Features
• Custom tokenization logic for flexibility
• Bidirectional LSTM to capture context from both directions
• Easy-to-train model using Keras Sequential API
• Clean, modular code for extensibility
📚 References
• Keras Documentation
• IMDb Dataset
• Bidirectional LSTM
