Data Science Capstone Project

The Coursera Data Science Specialization project (sponsored by SwiftKey) is to create an application that predicts the next word in a phrase/sentence. Such an application would be very useful for a keyboard on a mobile device that helps a user type faster by anticipating what the next word would be.

On the following slides, I will explain the details of the application created to fulfill the objective. We will go over:

• Overview of the Application
• Technical and Statistical Architecture
• Future Possibilities
• Conclusion and Appendix

The application uses text documents collected from blogs, news articles, and twitter to statistically model language patterns. N-Grams, Markov Model and Katz's back-off model were used to predict the next word. The modeling process used a large set of data so it was batched. The batch process created a much smaller set of data to be used in the application for fast real-time performance. The diagram below show the architecture of the application:

This project only scratches the surface of natural language processing and predictive analytics applications. There are many possibilities for improvement including:

• Parts of speech tagging to better understand the context of the user's words
• More data sources (such as books and Reuters Corpus Vol. 1)
• Parallel processing of data to increase volume and speed
• Feedback loop into the model (e.g., learn from what predictions users accept)

I found this project and the Data Science Program to be very enjoyable! I learned quite a bit and had fun doing it.

For this project, here is a list of resources:

• Github repository for the project
• Milestone report for the project
• Coursera course on NLP
• Text Mining Infrastructure in R
• Basic Text Mining in R
• tm Package in R
• Shiny Tutorial
• N-Grams