• Final project for the Johns Hopkins Data Science Capstone
• Goal: Predict the next word from a phrase using NLP
• Data: English blogs, news, and Twitter (HC Corpora)
• App deployed at:
  https://wishyouagoodfuture.shinyapps.io/capstone_shinyapp/

• Preprocessing:
  • Sampled ~1% from each source
  • Converted to lowercase, removed punctuation/whitespace
• Modeling:
  • Constructed unigram, bigram, trigram tables
  • Selected top 5000 TF-IDF words as vocabulary
  • Used backoff model: trigram → bigram → top tf-idf

Example outputs from our model:

• Predictions made using frequency + context match
• Efficient for real-time use in Shiny

• Simple UI: phrase input + predict button
• Fast prediction from preloaded .rds models
• Link:
  https://wishyouagoodfuture.shinyapps.io/capstone_shinyapp/

Try phrases like: - Let me know if you - She said she was going to

✅ Fast + lightweight model
✅ TF-IDF restricts vocabulary to high-signal words
✅ Handles unseen inputs via backoff fallback
✅ Shiny app loads fast and is interactive
✅ Deployed successfully and peer-review ready

Thanks for reviewing!