Documentation

The Dzongkha Next Word Prediction System is a machine-learning-based research initiative aimed at enhancing digital productivity and preserving the national language of Bhutan, through cutting-edge NLP and AI technologies.

System Guide Video

Detailed walkthrough of the predictive interface and syllable-based input methods.

Technical Methodology

1. Data Curation

Web scraping and compiling a diverse Dzongkha corpus from news, legal, and literary genres.

2. Syllable Segmentation

Utilizing rule-based segmentation verified by linguists for the Dzongkha script.

3. Model Training

Fine-tuning deep learning models using PyTorch and TensorFlow.

Development Stack

  • Languages: Python, JavaScript
  • AI Frameworks: PyTorch, TensorFlow
  • Web Architecture: Django Framework
  • Hardware: NVIDIA GPU for model training

Social Impact

🇧🇹

National Identity

Strengthening the presence of Dzongkha in the digital age.

⚡

Efficiency

Reducing typing effort by approximately 80%, boosting productivity.

🎓

Digital Literacy

Empowering students and educators with modern AI tools.