Next Word Predictor

Frank Hasbani
April/17/2015

Our purpose today is to introduce to you our predictive text model within the confines of the Capstone Project of Data Science Specialisation at JHU.

SwiftKey, the corporate partner in this capstone, builds a smart keyboard that makes it easier for people to type on their mobile devices. The cornerstone of their smart keyboard is predictive text models.

Presentation Summary

To showcase our Next Word Predictor:

  • We will briefly explain how our predictive model Algorithm is built
  • We will show the user how this data product works
  • We will review the architecture of the Next Word Predictor

The working Model is available for review URL: 'http://frankhasbani.shinyapps.io/final/'
Additional information about the preparation of the model URL: 'http://rpubs.com/FRH123/69391'

The Algorithm: How it works

We're applying the Markov Chain, using 2, 3 and 4grams, we smooth with the following Interpolation formula: Interpolation
And we apply the Katz Backoff with K = 3 KatzBackoff

How To Use It

Read the Instructions, Type in a word, partial or whole sentence, Click the Button, and Experience the predictions with more words Instructions

Under The Hood

Predictor The above caption is the Solution Architecture