Capstone Project: Next Word Prediction App

Problem and Goal

Typing on mobile devices can be slow and error-prone.
Our goal is to build an app that predicts the next word given a phrase to improve typing speed and accuracy.

Data and Preprocessing

• Data source: SwiftKey dataset (Blogs, News, Twitter)
• Cleaning steps:
  • Lowercasing, removing punctuation and numbers
  • Removing stopwords and profanity
  • Tokenization and stemming

Prediction Algorithm

We use a simple backoff n-gram model: - Try to match last 3 words → trigram - If not found, use last 2 → bigram - If still not found, use last word → unigram

About the App

• Built with R + Shiny
• Enter a phrase → predict next word
• Link: https://fturetto.shinyapps.io/NextWord

Summary and Future Work

• Shows potential of n-gram modeling
• Future: personalization, LSTM, BERT