09/01/2026
Typing on mobile devices is slow and error-prone. Modern keyboards solve this problem by predicting the next word a user intends to type.
The goal of this project is to build a next-word prediction algorithm and deploy it as a Shiny web application that can be used interactively.
This application demonstrates how statistical language models can improve user experience in everyday typing tasks.
The prediction algorithm is trained on a large English text corpus containing data from three sources:
Before modeling, the data were cleaned by: - Converting text to lowercase - Removing punctuation and numbers - Removing stop words - Tokenizing text into n-grams (1-grams, 2-grams, and 3-grams)
These preprocessing steps reduce noise and improve prediction accuracy.
The application uses an n-gram language model to predict the next word.
How it works: - The user inputs a phrase (one or more words) - The model looks for the most frequent next word based on previously observed word sequences - If no match is found, the model backs off to lower-order n-grams
This approach balances accuracy and speed, allowing the app to return predictions quickly.
The Shiny application provides a simple and intuitive interface:
The app runs on shinyapps.io and is accessible through a public URL. Predictions are generated dynamically based on user input.
The design emphasizes usability and fast response time.
The application provides accurate predictions for common English phrases from news and social media text.
User experience highlights: - Fast prediction response - Clean and minimal interface - Easy to use without technical knowledge
Future improvements could include: - Predicting multiple candidate words - Improving accuracy with larger datasets - Optimizing memory usage for scalability
This project demonstrates a practical and deployable data product.