Next Word Prediction App

Data Product Capstone Project

Kusuma

Prediction Algorithm

• Built using n-gram language models
• Uses bigrams and trigrams
• Trained on text from:
  • Twitter
  • News articles
  • Blogs
• Predicts the most frequent next word
• Falls back safely when no match is found

Shiny App Overview

• User enters a phrase (multiple words)
• App predicts one next word
• Prediction updates after clicking submit
• Deployed on shinyapps.io
• Accessible through any web browser

User Experience & Examples

• App responds quickly to user input
• Provides a prediction for every tested phrase
• Example inputs tested:
  • “one of”
  • “this is a”
  • “going to be”
  • “in the middle”
  • “as a result”
• Each input returned a valid next word

Conclusion & Value

• Demonstrates a practical NLP prediction system
• Simple, efficient, and easy to use
• Clear separation of model and interface
• Can be extended with larger datasets or smoothing
• Shows ability to build, deploy, and present a data product