author: Ayşe Çiğdem Şehitoğlu date: 18 October 2026

• Typing on mobile devices is slow and error-prone
• Users need fast and accurate next-word suggestions
• Goal: Build a lightweight text prediction app similar to SwiftKey

• A simple Shiny web application
• User types a phrase → App predicts the most likely next word
• Fast response and easy to use
• Built with R and deployed on shinyapps.io

• Cleaned text from blogs, news and Twitter
• Built n-gram models (bigrams and trigrams)
• Used Stupid Backoff strategy
• When a higher-order n-gram is not found, the model falls back to a simpler one
• Only the most frequent n-grams are kept to keep the model fast and small

1. Open the Shiny app
2. Type a phrase in the text box (example: “I want to”)
3. Click Submit or wait a moment
4. The app shows the predicted next word

Demo link will be provided in the final submission.

• Simple and understandable approach
• Fast enough for real-time use
• Easy to improve later (more data, better backoff, larger n-grams)
• Demonstrates end-to-end data science skills: data → model → product

Thank you