A real-time next word prediction app built with n-gram language modelling and Stupid Backoff. Trained on 102M+ words from blogs, news, and Twitter (HC Corpora). Deployed as an interactive Shiny app with sub-10ms prediction latency.

Exploratory data analysis of the HC Corpora dataset for the Johns Hopkins Data Science Capstone. Includes summary statistics, word frequency analysis, and plans for the n-gram prediction algorithm and Shiny app.