Word prediction

Text prediction Shiny App

    Coursera-Johns Hopkins University 
    Data Science Specialization
    Capstone Project 

Author: Simone Di Cicco

Date: September 17th, 2018

Word prediction

The goal is to develop a text predictive Shiny application which can predict the users next word with accuracy and speed.

The algorithm is based on a N-Gram model that was built from a large corpus provided by SwiftKey. Content is coming from:

  • Twitter
  • News
  • Blogs

Corpus attributes

Total Size Total Words
28MB 4,500,000

Shiny App User Interface

The UI designed to access the experience is very clean, simple and minimal.

Screenshot

The user enter text into the input box and gets the next word predicted.

The algorithm

The algorithm is looking first at accuracy level and then being enough fast to ensure a perfect user experience.

Trigram model

  • accurancy: 60%

  • speed: 0.5 to 1 second

Application Demo

Demo: ShinyApps