×

Fault-tolerant romanized input method for non-roman characters

  • US 7,810,030 B2
  • Filed: 10/03/2007
  • Issued: 10/05/2010
  • Est. Priority Date: 08/25/2004
  • Status: Active Grant
First Claim
Patent Images

1. A computer-implemented method for identifying most likely conversions of original inputs that include potentially incorrect original inputs, the method comprising:

  • receiving an original sequence of original Pinyin inputs, each original Pinyin input having associated proximity measurements, each proximity measurement being indicative of a probability of a respective intended Pinyin input being entered as the original Pinyin input;

    identifying each original Pinyin input that has proximity measurements indicating one or more non-zero probabilities for respective intended Pinyin inputs that are different from the original Pinyin input as a potentially incorrect original Pinyin input;

    generating alternative sequences of Pinyin inputs from the original sequence of original Pinyin inputs, each alternative sequence of Pinyin inputs including at least one intended Pinyin input that is different from a potentially incorrect original Pinyin input in place of the potentially incorrect original Pinyin input;

    determining a corresponding conversion to Hanzi characters for the original sequence of original Pinyin inputs and for each alternative sequence of Pinyin inputs;

    associating with each conversion a likelihood, each likelihood being based on the proximity measurements associated with the intended Pinyin inputs included in the corresponding sequences of Pinyin inputs;

    identifying most likely conversions for presentation, the most likely conversions including;

    a first conversion to first Hanzi characters converted from the original sequence of original Pinyin inputs, the first conversion having a first likelihood; and

    a second conversion to second Hanzi characters converted from one of the alternative sequences of Pinyin inputs and having a second likelihood that exceeds the first likelihood by a predetermined threshold.

View all claims
  • 2 Assignments
Timeline View
Assignment View
    ×
    ×