We built a lightweight predictive text model using R.
The system predicts the next word based on what the user types, using n-gram language modeling.
The final product is a user-friendly Shiny app deployed on shinyapps.io.

• Cleaned the text corpus
  
• Generated unigram, bigram, and trigram frequency tables
  
• Removed punctuation, stopwords, and numbers
  
• Stored the n-gram datasets in CSV format for fast lookup

Our backoff logic:

1. Check the last two words → trigram prediction
   
2. If no match → check last one word → bigram prediction
   
3. If still no match → return the most frequent unigram

Fast, deterministic, and lightweight.

• User types a phrase in the input box
  
• App runs the prediction model
  
• Displays the most probable next word
  
• Deployed on shinyapps.io

• Useful for keyboard suggestions, writing assistants, and chat apps
  
• Lightweight → no GPU required
  
• Extendable to advanced NLP / deep learning
  
• Shows real-world usage of R + Shiny + NLP