Word Prediction Application

These are the general steps taken to predict the next word:

• 5% of Twitter data & 1% of news data were used to build the model

• The model takes the user's input text and determines amount of words in the text, let's call it “n”

• Then the sampled data gets tokenized at n+1 grams

• The model then looks for matches of the input text to any lines in the n+1 grams tokenized data like this:

linesWithTheText <- tokenizedData[grepl(paste("\\<", inputText, "\\>", sep=""), text, ignore.case=T)]

• Once matches are found, the model select the ones that appeared in sampled data at highest frequency

• If there are ties, the model randomly selects one of them!

• Stop words were not excluded. These are good indicators for predicting the next word in common phrases!

• Input text next to “Your text:”
• Please avoid using puntuation marks as these were ignored in the model
• Please wait about 2 mins for the app to execute each job submitted
• Longer text takes alonger time. Try 2 words text input first
• Feature: same input text can be submitted more than once and the app may yield different answers!