This presentation explains a text prediction app for the Capstone project in the John Hopkins University Data Science specialization through Coursera and in partnership with Swiftkey.
How to use the app
How it works
Performance Summary
The App
The WhatsNext App demonstrates a text prediction algorithm. The app is available on shinyapps.
Instructions
Text is entered in the top box,
The three most likely next words are displayed
How the app processes text and calculates probabilities
Under The Hood
Determining the most likely next words
Clean incoming utterances
Match to n-gram frequency tables backing-off to lower order n-grams
Calculate probabilities
The table (right) shows matching the utterance: in the early part
The word of has the highest probability of being the next word
Performance
The user experience should be quick and responsive.
A balance between optimizing predictors and app processing time was found by measuring two processes,
