- Created in R Shiny, the application utilizes a custom input widget - Simply enter the text you want to predict the next word to - no further input is required.

- Leverages the top 1,000,000 most frequent 5-grams and 4-grams from the Corpus of Contemporary American English (COCA). - Lower order ngrams are derived from a corpus of US news article extracts provided by SwiftKey. - The selected subsets of the 5-grams and 4-grams maximize the frequency of the predicted word. - Lower order ngrams are subset on Kneser-Ney probability.

Accuracy was estimated by applying the algorithm to a 63 sentence-long New York Times article and assessing successful prediction rates of the 5th and last word of each sentence.

http://www.nytimes.com/2016/10/09/us/politics/donald-trump-campaign.html