• Do you need word prediction when you type in order to be more efficient?
• Do you want the predictions based on popular English words used in social media?
• Do you want an easy to use app?

If you answer ‘yes’ then this app is for you.

• Word Predictor uses the Softkey data set which includes tweet, news and blogs data.
• The data is used to train the N-Gram model. This version uses the Tri-Gram model.
• The Kneser-Ney Smoothing algorithm is used to rank the N-grams.
• When a user enters text into the app, the last two words are extracted to select the trigrams whose first two words match those words. The last words of those trigrams are the predicted values.

• If no trigrams are found, the last word of the text is used to select the bigrams whose first word is the same. The second words of those bigrams are the predicted values.
• If no bigrams are found, the unigrams with highest probabities in the data are chosen.
• The app displays the top 20 predicted words.

• The Word Predictor App can be found at https://minhvo.shinyapps.io/wordPredictor/
• Enter your text into the textbox of the app. The top 20 most likely words will display, five at a time, but this can be changed to ten or twenty, using the select box of the output table.
• The Documentation tab shows the instructions on how to use the app.