Data Science Capstone Project- SwiftKey Next word prediction

Uma Venkat
Jun 2017

What is it about?

The idea is predict the next possible word(s) for you to use while you type your sentence. So I built a model that will predict the next word in a sentence with relative accuracy. The model can form a sentence that is grammatically correct and with approppriate words.

Features of the application:

The application

  • Will predict words for start of a sentence and after every word in a sentence.
  • Will render appropriate capitalization of letters such as beginning or after a period, question mark or exclamation mark.
  • Will display contraction words appropriately.
  • Is child friendly in the sense that no profanity will be predicted.

What does the application do & How it works?

  • Load corpus (text(s))
  • If the corpus is large, break into smaller chunks. You can set the size of the data while calling the function.
  • Else clean the data. If the files are broken you may use the execclean function to process you smaller broken down files.
  • Once cleaned load you cleaned text file(s)
  • You can then use the exexgram function to produce the n-grams. By default you will have unigrams, bigrams & trigrams. If you want more then you can set it to a different value. The n-grams will produce a consolidated list of unique values even if you have multiple cleaned data files
  • Finally use pred function to predict the next word for your sentence. You ca set the number of results you want to get from the prediction.

How to use the application?

  • Choose from the list of words predicted for the start of a sentence
  • Type the word you choose and enter space for the next prediction
  • As you type the model will list words that you could use to finish the word
  • Choose from the list of words predicted for the following word in the sentence

Give it a try: https://ultraviolet.shinyapps.io/SwiftKey/

Finally

Thanks to