Word prediction app
Yoan Bidart
November, 6th, 2017
The project
This project have been made for the Coursera Data Science Specialization from John Hopkins University, the data has been provided by Swiftkey.
The aim of the project is to create a word prediction app, which can run on a portable device or a shiny app.
Basics of the app
The space necessary to run the app is 66 Mb, and this could not even slow down an old smartphone. It is actually running on Shiny App, at a fast pace.
The predicting algorythm whill calculate the best three words for your sentence, based on the previous 1 to 5 words (ngrams).
Let's talk accuracy
The accuracy is based on 50 000 predictions from our independent testing and validating sets.
As you can see the 5-grams accuracy is more than 10%.
To conclude
Getting a high accuracy on word predicting can be very hard, and needs a lot of computing power, not the ideal for a smartphone for example.
We think you will find a fast and small app for predicting words, and hope you will enjoy it.
Thank you for your attention!