🎭 Shakespearify

Transform modern English into Shakespearean style using state-of-the-art language models.

🎯 Overview

Shakespearify is an innovative text transformation tool that converts modern English into the eloquent style of Shakespeare. Using fine-tuned T5 models, it captures the essence of Shakespearean language while preserving the original meaning. The project features a Streamlit web app, robust post-processing, and a modular codebase for easy extension.

---

🛠️ Tech Stack

• Frontend/UI: Streamlit
• Modeling: HuggingFace Transformers, PyTorch
• Data Processing: Pandas, spaCy, NLTK
• Speech Synthesis: gTTS
• Testing: pytest, unittest.mock
• Visualization: Plotly
• Development: Jupyter Notebooks

---

🗂️ Project Structure & File Roles

shakespearify/
├── data/                # Dataset files (not in git)
├── img/                 # Project images and GIFs
├── notebooks/           # Jupyter notebooks for training and analysis
│   ├── shakespearify.ipynb
│   └── shakespearify-bart.ipynb
├── src/                 # Source code
│   ├── __init__.py
│   ├── app.py           # Streamlit web application (main UI & logic)
│   ├── postprocessing.py # Shakespearean text post-processing utilities
│   └── pages/
│       └── transformers.py # Interactive educational page on transformers & T5
├── tests/               # Test files (pytest)
│   ├── __init__.py
│   └── test_translate.py
├── t5-shakespeare/      # Fine-tuned model files (not in git)
├── t5-shakespeare2/     # Alternative model version (not in git)
├── requirements.txt     # Python dependencies
└── README.md            # Project documentation

File Descriptions

• src/app.py: Main Streamlit app. Handles language selection, model loading, translation, and speech synthesis.
• src/postprocessing.py: Cleans and enhances model output with phrase and word-level Shakespearean substitutions, using spaCy for POS tagging.
• src/pages/transformers.py: Streamlit page explaining transformer models and the T5 architecture interactively.
• notebooks/: Contains model training, evaluation, and exploratory analysis.
• tests/test_translate.py: Unit tests for translation logic using mocks.
• t5-shakespeare/, t5-shakespeare2/: Model checkpoints (not tracked in git; see below).

---

📊 Workflow & Architecture

graph TD
    A[Modern Text Input] --> B[Pre-processing]
    B --> C[BART Model]
    C --> D[Post-processing]
    D --> E[Shakespearean Text]
    E --> F[Speech Synthesis]
    
    subgraph Model Training
    G[Training Data] --> H[Fine-tuning]
    H --> C
    end
    
    subgraph Evaluation
    E --> I[BLEU Score]
    E --> J[ROUGE Score]
    E --> K[BERTScore]
    end

Loading

---

🚀 Features

• Convert modern English, French, or Spanish to Shakespearean English
• Interactive web interface with real-time translation and speech
• Phrase and word-level post-processing for authentic Shakespearean style
• Educational transformer model explorer
• Comprehensive evaluation metrics and test suite

---

📈 Model Performance

Metrics

• BLEU Score: 0.1453
• ROUGE Scores:
  • ROUGE-1: 0.4644
  • ROUGE-2: 0.2050
  • ROUGE-L: 0.4348
  • ROUGE-Lsum: 0.4347
• BERTScore F1: 0.6763

Sample Translations

1. Modern: "Indeed, and I am happy about that."

   Shakespeare: "Marry, And I am glad of it with all my heart:"

   Model: "Forsooth, and I am gleeful of that."

2. Modern: "What, by God, may she do? She may marry a king, a bachelor, a handsome young man. Certainly, your grandmother had a worse match."

   Shakespeare: "What, marry, may she? Marry with a king, A bachelor, a handsome stripling too. I wis, your grandam had a worser match."

   Model: "What, By my faith, may she dost? She may marry a Monarch, a bachelor, a handsome youthful gentleman. Certainly, thy grandmother had a worse match."

3. Modern: "That sounds good to me."

   Shakespeare: "It likes me well."

   Model: "That sounds well to me."

4. Modern: "I came as an ambassador from Edward, but I return as his sworn and deadly enemy. He told me to take care of his marriage, but he'll have war instead."

   Shakespeare: "I came from Edward as ambassador, But I return his sworn and mortal foe: Matter of marriage was the charge he gave me, But dreadful war shall answer his demand."

   Model: "I came as an ambassador from Edward, But I return as his sworn and fatally foe. He told me to take care of his marriage, but he 'll hast strife."

5. Modern: "I'll only irritate you if I stay. Let me go."

   Shakespeare: "I chafe you, if I tarry. Let me go."

   Model: "I will irritate thou if I stay. Let me hie."

---

🧑‍💻 Setup & Usage

1. Clone the repository

git clone https://github.com/MichAdebayo/shakespearify.git
cd shakespearify

2. Install dependencies

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

3. Download Model Files

Model checkpoints are not included in the repository due to size. Place your fine-tuned models files in t5-shakespeare/ and/or t5-shakespeare2/ as needed.

4. Run the Streamlit App

streamlit run src/app.py

---

🧪 Testing

Run the test suite with:

pytest tests/

---

📚 Educational Tools

• Explore transformer models and the T5 architecture interactively via the "How Transformers Work" page in the app (src/pages/transformers.py).

---

📄 License

MIT License

---

🤝 Contributing

Contributions are welcome! Please open an issue or submit a pull request.

---

🙏 Acknowledgments

• HuggingFace Transformers
• Streamlit
• spaCy, NLTK, gTTS
• Shakespearean datasets from Kaggle and other open sources

---

💡 Contact

For questions or suggestions, please open an issue on GitHub or contact the maintainer.