×

Method and apparatus for training a translation disambiguation classifier

  • US 7,318,022 B2
  • Filed: 06/12/2003
  • Issued: 01/08/2008
  • Est. Priority Date: 06/12/2003
  • Status: Expired due to Fees
First Claim
Patent Images

1. A method of training a classifier, the method comprising:

  • applying a first classifier to a first set of unlabeled data to form a first set of labeled data, the first classifier capable of assigning data to classes in a first set of classes;

    applying a second classifier to a second set of unlabeled data to form a second set of labeled data, the second classifier capable of assigning data to classes in a second set of classes that is different from the first set of classes; and

    using the first set of labeled data and the second set of labeled data to retrain the first classifier to form a retrained classifier that can be used to assign data to classes, retraining comprising;

    determining a first probability component from the first set of labeled data;

    determining a second probability component from the second set of labeled data; and

    combining the first probability component and the second probability component to form a probability term used to define the first classifier.

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