• This project demonstrates a next-word prediction app.
• Users input a phrase, and the app predicts the most likely next word.
• Built as part of the Data Science Capstone Project using the SwiftKey dataset.
• Goal: Mimic real-world mobile keyboard prediction.

• Data sourced from blogs, news, and Twitter (en_US).
• Steps taken:
  • Lowercasing, punctuation & number removal.
  • Profanity filtering.
  • Tokenization and n-gram (unigram to trigram) creation.
• Sampling used to reduce computational load.

• Uses n-gram language modeling (mainly bigram & trigram).
• Stupid backoff algorithm:
  • Try trigram → backoff to bigram → backoff to unigram.
• Predictions ranked by frequency of occurrence.
• Fast, lightweight, and interpretable.

• Built using R Shiny and hosted on shinyapps.io.
• Enter a phrase into the text box.
• The app returns the predicted next word.
• Real-time response or upon clicking submit.
• [https://j7f6j5-stephen0ayiah-fletcher.shinyapps.io/project_course/]

• The app demonstrates a basic NLP pipeline.
• Can be expanded with:
  • Deep learning (e.g., LSTMs or transformers)
  • User personalization
• Clean UI for non-technical users.
• [Insert RPubs slide link if needed]

Thank you!