Myanmar Next Syllable Prediction using LSTM

This project implements a deep learning approach to Next Syllable Prediction for the Myanmar language. Due to the unique script characteristics of Myanmar, this project utilizes a linguistically aware syllable-breaking strategy.

Demo - https://sanlinnaing.github.io/my-next-syllable-predict/

🛠 Technical Workflow

Phase 1: Linguistic Analysis and Syllable Extraction

Myanmar text lacks explicit word boundaries, requiring tokenization at the syllable level:

• Syllable Extraction: The syllable_base_pattern groups consonants and diacritics into single phonological units.
• Real-time Prediction Logic: The syllable_break_pattern isolates the Onset from the Rhyme, allowing suggestions even while mid-syllable.

Phase 2: Dataset Loading and Sequence Engineering

• Source: Data is sourced from the Wikimedia/Wikipedia Myanmar corpus.
• Sliding Window: A SEQUENCE_LENGTH of 5 syllables is used as context.
• Supervised Mapping: Generates pairs of Input (5 syllables) -> Target (1 next syllable) and partial syllable "tail" mappings.

Phase 3: Model Architecture

The model uses a Recurrent Neural Network (RNN) implemented in Keras:

• Embedding Layer: Maps syllables into a dense 256-dimensional vector space.
• Bidirectional LSTM: Captures forward and backward linguistic context.
• Mixed Precision: Utilizes mixed_float16 for faster training.

Phase 4: Inference and Generation

• Temperature Sampling: Adjusts the probability distribution for more or less "creative" results.
• Top-N Predictions: Returns multiple candidates, mirroring modern predictive text bars.

---

🏃 How to Run

It is highly recommended to use a Python Virtual Environment to manage dependencies.

1. Create and activate a virtual environment:

   # Create environment
   python -m venv venv
   
   # Activate (MacOS/Linux)
   source venv/bin/activate
   
   # Activate (Windows)
   .\venv\Scripts\activate

2. Install the required Python packages and Jupyter environment:

   pip install datasets mwparserfromhell seaborn scikit-learn tensorflow keras jupyterlab

3. Run Jupyter notebook

   jupyter notebook

License

This project is licensed under the GPL-3.0 License. See the LICENSE file for details.

Acknowledgement

it is the work working together with Gemini Code Assist.