Rank-reduced token representation for automatic speech recognition

  • US 10,593,346 B2
  • Filed: 03/15/2017
  • Issued: 03/17/2020
  • Est. Priority Date: 12/22/2016
  • Status: Active Grant
  • ×
    • Pin Icon | RPX Insight
    • Pin
First Claim
Patent Images

1. An electronic device, comprising:

  • a display;

    a microphone;

    one or more processors; and

    memory storing one or more programs configured to be executed by the one or more processors, the one or more programs including instructions for;

    receiving speech input via the microphone;

    determining a sequence of candidate words corresponding to the speech input, the sequence of candidate words including a current word and one or more previous words;

    determining, from a set of trained parameters, a vector representation of the current word, wherein a number of parameters in the set of trained parameters varies as a function of one or more linguistic characteristics of the current word, wherein a second vector representation of a previous word of the one or more previous words is determined from a second set of trained parameters, wherein one or more linguistic characteristics of the previous word is different from the one or more linguistic characteristics of the current word, wherein a number of parameters in the second set of trained parameters is different from the number of parameters in the set of trained parameters, and wherein a dimension of the second vector representation of the subsequent word is equal to a dimension of the vector representation of the current word;

    determining, using the vector representation of the current word, a probability of a next word given the current word and the one or more previous words; and

    displaying, based on the determined probability, a text representation of the speech input on the display.

View all claims