Predict-A-Tron 3000

Mary van Valkenburg
January 2016

We all need a helping hand at times

(Predict-A-Tron 3000 was created with that in mind)

P.A.T. 3000 predicts the next word as you type, using the Markov assumption that future events depend only on a limited history of preceding events.

P.A.T. 3000 is an n-gram model

built from large text files that were harvested from twitter, news posts, and blogs.

bigrams: “argh why”, “why do”, “do I”, “I have”, “have to”, “to do”, “do everything”, “everything myself” trigrams: “argh why do”, “why do I”, “do I have”, “I have to”, “have to do”, “to do everything”, “do everything myself” fourgrams: “argh why do I ”, “why do I have”, “do I have to”, “I have to do”, “have to do everything”, “to do everything myself”

Step by Step

  • Tables of bigrams, trigrams, and fourgrams were created with computed counts for each entry that appears more than once in the combined corpora.
  • Applying Katz's backoff algorithm, P.A.T. 3000 looks for matches in the three n-gram tables (looking first to the fourgrams, then the trigrams, and finally the bigrams) and offers up the most probable next word when it is found.

Try Predict-A-Tron 3000 Today!

(because you don't have to do everything yourself) https://marylvv.shinyapps.io/PredictATron3000/