Challenge: Predict the next word as you type (like smartphone keyboards)

Our Solution:

• N-gram language models (2-5 grams)
• Trained on Coursera-SwiftKey dataset
• Fast backoff algorithm
• Simple Shiny web app

Algorithm:

Input: "I want to"
  ↓
Try 4-gram: "I want to ___"
  ↓ (not found)
Try 3-gram: "want to ___"
  ✓ Found predictions!
  ↓
Return: "go", "see", "be", "know", "do"

Key Idea: Start with longest context, fallback to shorter

Features:

• Type your text
• Click “Predict”
• Get 5 suggestions
• Shows which model was used

Try it: Launch App

Summary:

• Built working word prediction app
• Uses n-gram models with backoff
• Fast and memory-efficient
• Ready for deployment