2026-01-01

Problem Statement

  • Typing on mobile devices is slow and error-prone
  • Predicting the next word improves speed and user experience
  • Goal: build a simple and fast next-word prediction app

Data & Model

  • Based on N-gram language models
  • Uses:
    • Unigrams (single words)
    • Bigrams (two-word phrases)
    • Trigrams (three-word phrases)
  • Higher-order n-grams are given priority

Prediction Algorithm

  • User enters a phrase
  • Algorithm checks:
    1. Trigram matches
    2. Bigram matches
    3. Most frequent unigram fallback
  • Designed for speed and simplicity

Shiny Application

  • Interactive web interface using Shiny
  • Real-time next-word prediction
  • Easy-to-use text input
  • Lightweight and responsive

Conclusion

  • Demonstrated a working prediction model
  • Successfully deployed using Shiny
  • Can be extended using larger datasets
  • Suitable for real-world text prediction systems