Method and apparatus for training a translation disambiguation classifier
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.
2 Assignments
0 Petitions
Accused Products
Abstract
A method of training a classifier includes applying a first classifier to a first set of unlabeled data to form a first set of labeled data. The first classifier is able to assign data to classes in a first set of classes. A second classifier is applied to a second set of unlabeled data to from a second set of labeled data. The second classifier is able to assign data to classes in a second set of classes that is different from the first set of classes. The first and second sets of labeled data are used to retrain the first classifier.
34 Citations
31 Claims
-
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. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14)
-
-
15. A computer-readable medium having computer-executable instructions for performing steps comprising:
-
generating first language labeled data that indicates a sense of at least one word in a first language;
generating second language labeled data that indicates a sense of at least one word in a second language; and
using the first language labeled data and the second language labeled data to train a classifier for the first language, where the classifier can be used to identify a sense of a word in the first language. - View Dependent Claims (16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31)
-
Specification