This project presents an interactive web application that predicts the next word based on a user-entered phrase using a statistical Natural Language Processing (NLP) approach. The model is built using an n-gram language model, where word sequences are analyzed to learn patterns of word co-occurrence from a training text dataset. The application processes user input in real time by cleaning and tokenizing the text, then applying a backoff strategy using unigram, bigram, and trigram frequency tables to generate the most likely next word. If a higher-order match is not found, the model falls back to lower-order n-grams to ensure a prediction is always produced. The app is implemented in R using the Shiny framework, providing a simple and interactive interface where users can type a phrase and instantly receive a predicted next word. This project demonstrates the basic principles behind modern autocomplete systems used in messaging apps, search engines, and predictive keyboards.

This report presents an exploratory analysis of blogs, news, and Twitter data from the SwiftKey dataset. It summarizes key statistics, visualizations, and outlines a plan for building a next-word prediction model and Shiny application.