Entropy-guided text prediction using combined word and character n-gram language models

  • US 10,078,631 B2
  • Filed: 05/15/2015
  • Issued: 09/18/2018
  • Est. Priority Date: 05/30/2014
  • Status: Active Grant
  • ×
    • Pin Icon | RPX Insight
    • Pin
First Claim
Patent Images

1. A method for determining word prediction candidates to be displayed, the method comprising:

  • at an electronic device;

    receiving a first typed character from a user;

    determining a first entropy of a first set of possible word completions based on first probabilities of the first set of possible word completions, wherein the first probabilities are based on the first typed character;

    receiving a second typed character from the user;

    determining a second entropy of a second set of possible word completions based on second probabilities of the second set of possible word completions, wherein the second probabilities are based on the first typed character and the second typed character;

    determining a reduction in entropy from the first entropy to the second entropy, wherein determining the reduction in entropy comprises;

    determining the reduction in entropy based on third probabilities of a third set of possible word completions, the third set of possible word completions comprising words in the first set of possible word completions other than words in the second set of possible word completions; and

    in response to the reduction in entropy exceeding a threshold, causing a candidate word to be displayed from the second set of possible word completions.

View all claims