Problem & Motivation

• Typing on mobile devices can be slow and error-prone.
• Next-word prediction improves typing speed and accuracy.
• This app simulates a basic predictive text engine using public language data.

Algorithm & Model

• Built using n-gram models (unigram, bigram, trigram).
• Data cleaned: stopwords, numbers, profanity removed.
• Stupid backoff approach to choose highest probability next word.

App Functionality

• Enter any phrase (2-3 words) in the input box.
• App returns a predicted next word based on trained n-gram frequency tables.
• Hosted at shinyapps.io for live demo.

Insights & Next Steps

• Accurate for common phrases and social media style text.
• Plans: Expand n-gram depth, integrate deep learning models (e.g., LSTM).
• Try the app: https://prathamboricha.shinyapps.io/wordpredictor

Thank You

• Questions?
• Feedback welcome!