A Shiny App for Predicting the Next Word in a Phrase

• Smart keyboards and messaging apps often suggest the next word
• Useful for:
  • Improving typing speed
  • Assisting accessibility tools
  • Autocomplete features

Goal: Build a model that can predict the next word from a given phrase using real-world language data.

Data Source: - SwiftKey Dataset - Text from blogs, news, and Twitter

Method: - Tokenize text into Bigrams - Clean and normalize input - Match user input to known patterns

Prediction Logic:

```r predict_next_word <- function(input, ngram_data_list) { # match user phrase to bigrams }

🖥 The App (Demo Instructions)

• Built with Shiny
• Loads cleaned N-gram data
• Input: Any phrase (e.g. “hello”)
• Output: Predicted next word

1. Enter a phrase (at least one word)
2. Click “Predict Next Word”
3. View the predicted result

👉 Hosted on shinyapps.io

Performance: - Predicts common word pairs well - Example: - Input: "hello" → Output: "world"

Next Improvements: - Add Trigram & Quadgram support - Implement Katz Backoff or Stupid Backoff - Allow mobile deployment

Final Thoughts

• Simple but effective model