Jeremy Beck
January 2015
JHU's Coursera Data Science Specialization Capstone Project
Small keys in mobile keyboards make typing difficult. If we can predict the next word a user wants to type, they can simply select the word, without having to type the entire thing. To prototype this functionality, I designed an application that, given user input, predicts the next word in the phrase.
The requirements for this usage suggest that the app should:
For the purpose of this project, shinyapps.io was used to deploy the app.
The app uses a 30% sample of Twitter posts, News articles, and Blog posts as a basis for predictions, and N-grams ranging from unigrams to trigrams. The processing pipeline and data are available from an earlier report
The prediction algorithm is an implementation of Modified Kneser-Ney Smoothing/Interpolation for trigrams.
See the app in action Thanks for looking!