Swiftkey Word Predictor

An Overview by Z Ritchken

Overview

  • How it works:
  • The app is based on a predictive algorithm that uses the Katz Back-Off Model.
  • Estimates likelihood of an "n-gram" given the history of its evidence within the data.
  • An "n-gram" is a string of 1, 2, 3, or 4 words in a row.
  • The data that formed the basis of the data comes from Blogs, Twitter, and News stories.

How to use the App

  • In the field on the left side of the app, type a few words, generally a normal English phrase.
  • As you type, the app will predict the most likely next word, which appears in yellow. Other likely words appear in green. works

Try it out!

  • Below is an article from the New York Times from August 16, 2015. Try typing in some of the highlighted sentences and see what the app predicts:

trump

Additonal Uses

  • The underlying technology here exists in many smart devices already to allow prediction of messages.

  • A futher application could be translation; matching common strings of words of languages together to suggest what the user wants to say.

  • Thanks to Coursera and Swiftkey for providing the data and guidance!

    coursera swift