--- title: "Text Prediction App" author: "Anila Mohanan" output: ioslides_presentation --- ## Introduction - Built as part of Coursera Data Science Capstone - Predicts the next word in a user-entered phrase - Uses blogs text dataset is used. --- ## How the Algorithm Works - Text cleaned and tokenized using `tidytext` - Created **unigram**, **bigram**, and **trigram** frequency tables - Uses **n-gram model with backoff**: - If trigram not found → backoff to bigram → backoff to unigram --- ## Prediction Model - Efficient and memory-optimized using RDS storage - Fast lookup using `dplyr` filtering and `slice_max` - Handles unknown inputs with fallback strategy --- ## Shiny App Demo - Built using **Shiny** and hosted on **shinyapps.io** - User enters a phrase → App predicts the next word - Try it live here: **https://x1tkzd-anila-mohanan.shinyapps.io/NextWordPredictor/** --- ## Summary - Accurate, fast predictions with minimal resources - Real-time prediction from cleaned social and web text - Could be extended into mobile keyboards or chat assistants

This report presents a summary of the initial steps in developing a text prediction model using n-grams. It includes data loading, cleaning, tokenization, and a strategy for building a Shiny app.