Data Science Capstone Final Project Submission

This presentation is a part of the Data Science Capstone delivered by the John Hopkins University on Coursera. The aim of the course project is to develop a Shiny application using natural language processing. The documentation on this project is as follows:

• The GitHub repository containing all the source codes for this project (ui.R, server.R and this presentation Capstone_project.Rpres): https://github.com/elfedorova/10_Data_Science_Capstone_Project
• The deployed shiny application is available by the link: https://elfedorova.shinyapps.io/Word_predictor/
• The source data: https://d396qusza40orc.cloudfront.net/dsscapstone/dataset/Coursera-SwiftKey.zip

The Shiny application of our interesest represents a tool that can predict the next word, like the one used in mobile keyboards applications implemented by the Swiftkey.

Here, the user enters the text in an input box, and in the other one, the application returns the most probable word to be used. While entering the text, the field with the predicted next word refreshes simultaneously, and then the predicted word is provided.

The predicted word is obtained from the n-grams matrices, comparing it with tokenized frequency of two, three and fourgram sequences. Additionally, the panel indicates the type of a matrix that was applied in the prediction.

• The source data consists of four chunks (relative to the language of a data set). For the purpose of the project we use en_US data sets.

• Having loaded the data, we can perform cleaning: conversion to lower case, removal of the punctuation, whitespace, numbers and stopwords. The corpus was converted to plain text.

• As a next step, the tokenizer was created in order to construct the prediction models.

• Ultimately, the tables of n-grams (unigram, bigram, trigram and fourgrams) are used to predict the next word based on the text entered by the user.

In the suggested GUI the user inputs a phrase or a word into the textbox, and the application throws the most likely next word based on the word frequency, identifying the exact type of a matrix that was used in the particular case.