Text Prediction App

Introduction

This Text Prediction Application is a submission for the Coursera Data Science Specialization in collaboration with the Johns Hopkins University, Data Science Capstone module.

The inspiration of this application came from the SwiftKey Keyboard application, which is a type of text predicting tool that is capable of predicting the next word as user types onto the smart phone keyboard.

Behavior

The model used for this text prediction is by using the Ngrams, by default this application predicts three words. For this job it was necessary to build from unigrams to heptagrams. Preference is given to the longer sentences of the input text (a sentence has a maximum of 6 words), if it is not found in the longer ngrams, the sentence is reduced by word until it matches in the lower ngrams.

Application

This image shows the Text Prediction App in running!

Improvements

The prediction model is somewhat skewed. Possible improvements are,

• Increase Corpus size and/or type.
• Increase machine resources, create more complex model.
• Increase Shiny io resources, host more complex text prediction application.

Links

Application (Shinny): https://luxito88.shinyapps.io/TextPredictionApp/

Slides Pitch (Rpubs): http://rpubs.com/Luxito88/PitchTextPredApp